version:8.10.0-alpha1-rc2

Bug Fixes

Misc

• Regression in 8.9.1 leads to variables local to an ad-hoc sub-process inner instance scope to leak into the parent scope (#51939)

version:8.10.0-alpha1-rc1

Enhancements

Java Client

• API: Allow filtering and sorting by BusinessId in process instances search queries (#47246)
• Missing Exponential Backoff for OIDC Token Requests in Camunda Java Client (#44815)
• Use OffsetDateTime for date-time fields in java client (#33678)

Misc

• Engine: Propagate Business Id to child process instances (Call Activity: inherit-only for MVP) (#44980)
• Gateway: Stable partition routing for Business Id uniqueness (#44979)
• API: Support Business Id on process instance start (REST + gRPC) (#44977)
• Optimize: support extraConfiguration via SPRING_CONFIG_IMPORT (parity with orchestration) (#44820)
• [REST endpoints] Forward compatibility (#41936)
• Improve backup/restore usability: always ensure missing index templates are created (>=8.8) (#40675)
• [CPT] I can use a test environment globally (#29131)
• Support client side load balancing (#9870)

Bug Fixes

Broker

• Fail to throw/catch error if the job is based by a message throw event (#8662)
• RoleChangeListener is not called after becoming Leader (#4409)

Gateway

• NullPointerException in MCP gateway for null userTaskKey in getUserTask (#48201)
• NullPointerException in MCP gateway for null userTaskKey in searchUserTaskVariables (#48197)
• NullPointerException in MCP gateway for null incidentKey in resolveIncident (#48196)
• Partition retry inconsistency between gRPC and REST gateways for process instance creation (#46756)
• Gateway can easily overwhelm broker with handling blocking jobs (long polling) (#7659)

Java Client

• [CPT] Slow test execution and test failures if the client prefers REST (#45667)

Misc

• Tests for Swagger consistently fail for Legacy Operate E2E CI (#51702)
• Navigating to deleted process from Batch Operation view leads to blank screen (#51530)
• Variables tab stuck in infinite loading state when editing or adding a large (truncated) variable (#51469)
• Uncaught error incidents not visible in Operate process diagram view (#51400)
• Process instance migration falsely rejects mapped catch events on multi-instance subprocesses (#51295)
• Partition blocked by endless RocksDB retry loop Caught recoverable exception (#50958)
• fix: MCP tool schemas emit Swagger sentinel ##default as default value (#50934)
• Millions of logs in load tests that subscription can't be found (#50655)
• ZeebeDbInconsistentException on GLOBAL_LISTENER_PINNED_CONFIG (#50518)
• Tasklist pod OOM crash when opening advanced filter modal with 1000+ process definitions (#50396)
• SUSPEND audit log entityKey does not match batchOperationKey (#50311)
• Operate Operations Log filters show Identity-related entities (#50305)
• Optimize user task import spams logs for missing due date (#50145)
• Incident panel remains visible (shows 0 incidents) when navigating to a task without incidents (#50002)
• Scheduled flush timer in CamundaExporter get starved under sustained load (#49820)
• Failure to deserialise DbCheckpointState (#49812)
• NullPointerException for OperationType when handling BatchOperationChunkRecordValue (#49765)
• Task is not highlighted after navigating to History tab (#49707)
• fix: e2e job API tests share job type across parallel workers causing flaky 409 tests (#49604)
• Audit Log API: Sentinel Values Returned Instead of null (#49564)
• per-process-definition round-robin dispatch for CreateProcessInstance (#49530)
• Bug: Operate Listeners tab shows all jobs instead of only execution/task listeners (#49525)
• RESUME audit log entityKey does not match batchOperationKey (#49513)
• Inconsistent display: Identity name shows as 'Identity' in dashboard/app switcher but as 'Admin' in cluster details (Helm Chart, 8.9 snapshot) (#49512)
• Stored External Content Injection (#49491)
• Missing Navigation to Root Decision Instance in DMN Incidents (#49425)
• Batch Operation Details page cannot be scrolled (#49410)
• Incident tab context remains visible after incident resolution in Operate details view (#49220)
• Make incidents tab incident counter respond to current user selection (#49161)
• Swagger UI POST requests results in a 401 with CSRF validation error (#48902)
• Authorization search endpoint does not return page.startCursor and page.endCursor (#48857)
• Assigning roles to groups during initialization doesn't work (#48856)
• USER_TASK authorization cannot be re-created (#48847)
• Fullscreen toggle does not work in Operate diagram controls (Chrome) (#48823)
• UI: No margin in Input/Output Mappings panel list (#48801)
• Process instance details page doesn't handle permission errors on process instance history (#48728)
• MCP gateway: invalid search filter values cause 500 with Java class name leak (#48578)
• Batch move of regular ad-hoc subprocess corrupts the process instance (#48530)
• Operate UI shows incorrect token count when moving MI instances to an outside element (#48527)
• Process Instance with tag does not complete (#48502)
• duplicate WHERE clause in MessageSubscriptionMapper getProcessDefinitionStatistics query (#48387)
• Tooltip not displaying correctly when hover over on the 3 dots for batch operation details page (#48386)
• Admin Operations Log. Filters are using different styles and some filters can't be reset (#48279)
• Client Entity Incorrectly Labeled as "User" in Admin Operations Log (#48256)
• Swagger UI again only works using the authenticated tab (#48243)
• Task History tab content is not updated after some operation was perform (#48240)
• Excessive v2/element-instances/search requests and Instance History flickering on navigation (#48198)
• NullPointerException in MCP gateway for null incident key parameter (#48195)
• Optimize-client: Make maven build resilient to 'yarn install' failures (#48153)
• Raft bootstrap can block broker shutdown (#48136)
• Persistent web sessions broken on H2 - CAST(? AS BLOB) type mismatch (#48092)
• Resource access check failing when retrieving specific Global Task Listeners via API (#48056)
• Unauthenticated OIDC API requests return an invalid OAuth 2.0 Protected Resource Metadata link (#47773)
• [CPT] Process coverage report doesn't show completed ad-hoc sub-process elements (#47653)
• Optimize FE crashes with invalid dashboard filters ([#47595](https://github.com/camunda...

version:8.9.2

Bug Fixes

Java Client

• [Java client] Topology response doesn't return the property lastCompletedChangeId (#51559)

Misc

• Regression in 8.9.1 leads to variables local to an ad-hoc sub-process inner instance scope to leak into the parent scope (#51939)
• Tests for Swagger consistently fail for Legacy Operate E2E CI (#51702)
• Navigating to deleted process from Batch Operation view leads to blank screen (#51530)
• Process instance statistics API does not return incident counts for multi-instance bodies correctly (#51251)
• Partition blocked by endless RocksDB retry loop Caught recoverable exception (#50958)
• GET Decision Instance API doesn't return Bad Request (#43425)
• Some endpoints are falsely disabled when secondary storage is disabled (#40199)

Maintenance

• Spring Client stream warning (#25535)

Task

• Bump thymeleaf version (#51487)
• load-tests: remove backward compatibility for single-zone scheduling (#50568)
• Add CVE scanning to frontend runtime deps (#50406)
• Add comprehensive e2e ITs for the full archiver pipeline (#49826)
• Prevent NPE when permissions configuration has null/empty value (#42323)
• [Improvement] Automate creating the upgrade plan for Optimize release (#40258)
• Reduce log level of client timeouts warning for job stream (#31577)

Merged Pull Requests

• feat: add streamInactivityTimeout watchdog to JobStreamer (#51616)
• feat: opt-in OIDC userinfo augmentation for bearer tokens (#51537)
• fix: upload GCS backup files via InputStream to avoid CRC32C mismatch (#45646)
• fix: several user endpoints do not require secondary storage (#40200)
• fix: schedule jobs stream recreation (#35892)
• deps: Update spring boot to v3.3.0 (main) (#18776)

version:8.8.23

Bug Fixes

Broker

• Fail to throw/catch error if the job is based by a message throw event (#8662)

Misc

• Uncaught error incidents not visible in Operate process diagram view (#51400)
• Process instance migration falsely rejects mapped catch events on multi-instance subprocesses (#51295)
• Millions of logs in load tests that subscription can't be found (#50655)
• Incidents do not get handled correctly on job worker to Camunda user task migration (#49714)
• 'Assign to me' spinner blocks assignment when switching tasks quickly (#49133)
• Unassign User Task with wrong error description on 404 (#38880)
• All values in a nested variable are merged to parent scope when one value from the variable is listed as an output mapping (#35251)

Task

• Bump thymeleaf version (#51487)
• Ensure all release load tests run on single AZ with stable VMs (#50877)
• Improve event applier golden files experience (#36434)
• Investigate better default bulk delay for Camunda Exporter (#33165)

Merged Pull Requests

• feat: opt-in OIDC userinfo augmentation for bearer tokens (#51537)
• fix(operate): use unique process ID to fix flaky internal API permission tests (#51365)

version:8.9.1

Enhancements

Java Client

• Missing Exponential Backoff for OIDC Token Requests in Camunda Java Client (#44815)

Bug Fixes

Broker

• Fail to throw/catch error if the job is based by a message throw event (#8662)

Java Client

• [CPT] Slow test execution and test failures if the client prefers REST (#45667)

Misc

• Uncaught error incidents not visible in Operate process diagram view (#51400)
• Process instance migration falsely rejects mapped catch events on multi-instance subprocesses (#51295)
• Partition blocked by endless RocksDB retry loop Caught recoverable exception (#50958)
• fix: MCP tool schemas emit Swagger sentinel ##default as default value (#50934)
• Millions of logs in load tests that subscription can't be found (#50655)
• Actor Icon Same for User and Client in Admin Operations Log (#50598)
• ZeebeDbInconsistentException on GLOBAL_LISTENER_PINNED_CONFIG (#50518)
• Tasklist pod OOM crash when opening advanced filter modal with 1000+ process definitions (#50396)
• SUSPEND audit log entityKey does not match batchOperationKey (#50311)
• Operate Operations Log filters show Identity-related entities (#50305)
• Optimize user task import spams logs for missing due date (#50145)
• Evaluate an expression API response schema expects result to be unknown, not null (#50091)
• io.camunda.zeebe.engine.processing.processinstance.ProcessInstanceMigrationMigrateProcessor$SafetyCheckFailedException: Expected to migrate a job for user task but could not find job with key '0' (#50023)
• Usage metrics don't track RPIs started via message (#50010)
• Incident panel remains visible (shows 0 incidents) when navigating to a task without incidents (#50002)
• Failure to deserialise DbCheckpointState (#49812)
• Standalone Decision Instance Archiver job fails constantly (#49761)
• Incidents do not get handled correctly on job worker to Camunda user task migration (#49714)
• Task is not highlighted after navigating to History tab (#49707)
• Audit Log API: Sentinel Values Returned Instead of null (#49564)
• Bug: Operate Listeners tab shows all jobs instead of only execution/task listeners (#49525)
• RESUME audit log entityKey does not match batchOperationKey (#49513)
• Inconsistent display: Identity name shows as 'Identity' in dashboard/app switcher but as 'Admin' in cluster details (Helm Chart, 8.9 snapshot) (#49512)
• Batch Operation Lifecycle Audit Log Entries Are Not Deduplicated (#49428)
• Missing Navigation to Root Decision Instance in DMN Incidents (#49425)
• Batch Operation Details page cannot be scrolled (#49410)
• Operations Log table. 'Version' filter resumes other filters (#49391)
• Make incidents tab incident counter respond to current user selection (#49161)
• 'Assign to me' spinner blocks assignment when switching tasks quickly (#49133)
• Failed to start with Spring AOT cache on (#49131)
• Assigning roles to groups during initialization doesn't work (#48856)
• UI: No margin in Input/Output Mappings panel list (#48801)
• Schema Manager can fail reads and deletes with long requests (#47023)
• [8.9.0-alpha4+] Legacy Operate OpensearchConnector ignores awsEnabled — crashes on IRSA deployments (#45761)
• Optimize cannot output JSON logs with Stackdriver (#45223)
• Cannot assign user/group/tenant/etc. with slashes in the name (#45215)
• [CPT] Allow setting preferRestOverGrpc=true in CamundaSpringCamundaProcessTest (#45177)
• Leftover service worker rewrites URLs of new Identity UI (#44867)
• Operate V1 decision instance search fails for standalone evaluated decisions (#44592)
• Scrolling gets stuck in Tasklist v1 with large process list and search returns 500 error (#44583)
• Tasklist filter resets after selecting completed task (#43889)
• Elasticsearch unreachable during Zeebe startup and shutdown: exporter retries, timeouts, and thread interruption errors (#43266)
• Call hierarchy is not correct on modified process instance (#42729)
• Stop the OC when port 26502 is already in use (#42581)
• Batch Operations show 'Completed' state even when operationsCompletedCount !== operationsTotalCount (#42478)
• Bug: Tenant cluster variable creation responds 200 OK for invalid tenant instead of 404/400 (#42049)
• Handle WebSessionRepository exception when failing to save session (#39705)
• Make archiver bypass delayBetweenRuns when backlog exists; reschedule next run immediately (#39196)
• No error message is returned in UI when a user without Batch Operation Authorization try to migrate process instance on SM env (#39075)
• Unassign User Task with wrong error description on 404 (#38880)
• Sort by Version fails after applying Process Instance filter (#38103)
• When user has no access to update userTask no message is displayed in the UI (#37230)
• All values in a nested variable are merged to parent scope when one value from the variable is listed as an output mapping (#35251)
• Release: Maven Release Plugin commits fail spotless check due to XML formatting inconsistency (#35247)
• Incident metrics are greater than the actual number of incidents (#34328)
• java.lang.IllegalStateException: Expected to find element instance for given key, but didn't exist. (#26755)
• Scheduled commands cache uses the wrong intent for caching message expiry (#15574)
• Error on ProcessInstanceCreation with large variable size (#13016)
• When submitting a payload lower than the configured maxMessageSize, the request fails with EXCEEDED_BATCH_RECORD_SIZE (#12873)
• Create process instance with start instructions exceeding record batch size leads to errors (#10336)

Maintenance

• CI: move all repository secrets to Vault (#18211)

Task

• Optimize load test Docker image builds (#50881)
• Ensure all release load tests run on single AZ with stable VMs (#50877)
• Fix Camunda Exporter nightly tests on AWS (#50755)
• Fix flaky ElasticsearchPayloadSizeIT.shouldAcceptOrRejectBasedOnPayloadSize (#50739)
• [CPT] Flaky test io.camunda.ConnectorProcessTest.shouldInvokeConnectors (#50530)
• Passing correct tags for components in release load test (#50323)
• Schedule load tests in a single availability zone (#50028)
• Add missing unit test for flush timer (#50003)
• Better handle nested incidents on the new incidents tab ([#49165]...

version:8.7.28

Bug Fixes

Misc

• Process instance migration falsely rejects mapped catch events on multi-instance subprocesses (#51295)
• Partition blocked by endless RocksDB retry loop Caught recoverable exception (#50958)
• Millions of logs in load tests that subscription can't be found (#50655)
• 'Assign to me' spinner blocks assignment when switching tasks quickly (#49133)
• All values in a nested variable are merged to parent scope when one value from the variable is listed as an output mapping (#35251)
• java.lang.IllegalStateException: Expected to find element instance for given key, but didn't exist. (#26755)

Task

• Optimize load test Docker image builds (#50881)
• Ensure all release load tests run on single AZ with stable VMs (#50877)

Merged Pull Requests

• fix: reset AssignButton state when switching tasks to prevent spinner blocking assignment (#50988)

version:8.8.22

Bug Fixes

Misc

• Partition blocked by endless RocksDB retry loop Caught recoverable exception (#50958)
• Scroll contexts leaked in ElasticsearchUtil.scrollWith on error paths (#50892)
• Tasklist pod OOM crash when opening advanced filter modal with 1000+ process definitions (#50396)
• Date picker inputs are not populated correctly with existing dates (#50378)
• Optimize user task import spams logs for missing due date (#50145)
• io.camunda.zeebe.engine.processing.processinstance.ProcessInstanceMigrationMigrateProcessor$SafetyCheckFailedException: Expected to migrate a job for user task but could not find job with key '0' (#50023)
• Usage metrics don't track RPIs started via message (#50010)
• Post exporter queue reads from beginning after upgrade to 8.8 (#49374)
• Failed to start with Spring AOT cache on (#49131)
• Assigning roles to groups during initialization doesn't work (#48856)
• Optimize-client: Make maven build resilient to 'yarn install' failures (#48153)
• Schema Manager can fail reads and deletes with long requests (#47023)
• Cannot assign user/group/tenant/etc. with slashes in the name (#45215)
• Scrolling gets stuck in Tasklist v1 with large process list and search returns 500 error (#44583)
• Call hierarchy is not correct on modified process instance (#42729)
• Stop the OC when port 26502 is already in use (#42581)
• Batch Operations show 'Completed' state even when operationsCompletedCount !== operationsTotalCount (#42478)
• Handle WebSessionRepository exception when failing to save session (#39705)
• Make archiver bypass delayBetweenRuns when backlog exists; reschedule next run immediately (#39196)
• No error message is returned in UI when a user without Batch Operation Authorization try to migrate process instance on SM env (#39075)
• Progress bar remain appear after retry in operate (#38518)
• QueryBatchOperationTest is flaky (#37289)
• Operate process instance modification only cancels 10 tokens in a multi instance (#37282)
• When user has no access to update userTask no message is displayed in the UI (#37230)
• java.lang.IllegalStateException: Expected to find element instance for given key, but didn't exist. (#26755)
• Scheduled commands cache uses the wrong intent for caching message expiry (#15574)

Maintenance

• CI: move all repository secrets to Vault (#18211)

Task

• Optimize load test Docker image builds (#50881)
• Fix flaky ElasticsearchPayloadSizeIT.shouldAcceptOrRejectBasedOnPayloadSize (#50739)
• [CPT] Flaky test io.camunda.ConnectorProcessTest.shouldInvokeConnectors (#50530)
• Passing correct tags for components in release load test (#50323)
• Schedule load tests in a single availability zone (#50028)
• In daily stress tests use both REST and gRPC for comparison (#48977)
• Run with REST API by default in long-running load tests (#48976)
• Change default load test setup to include full coverage (main) (#48742)
• Add camunda-process-test-spring-boot-3 to 8.8 (#48283)
• Process instance archiving reindexes already archived process instances (#48200)
• Rename camunda-process-test-spring-4 to camunda-process-test-spring-boot-4 (#47483)
• OAuth2AuthenticationException are logged at ERROR in gateway (#46234)
• Refactor archive date tracking logic in Elasticsearch and Opensearch archiver repositories (#43931)
• (P1) Re-enable and validate accessibility color-contrast checks (#40261)
• Handle WebSessionRepository exception when failing to save session (#39705)

Merged Pull Requests

• test: fix failing ProcessZeebeImportIT (#50914)
• fix: always clear ES scroll context on exception (#50896)
• test: Make io.camunda.ConnectorProcessTest.shouldInvokeConnectors reliable (#50548)
• fix: resolve flaky UsageMetricRpiTest by waiting for state commit before assertions (#50467)
• fix: align OpenSearch load test config with Elasticsearch settings (#50463)
• fix: gracefully handle ES/OS failures in health indicators (#50432)
• refactor: replace use-official-docker-images with orchestration-tag and add per-component image tags (#50417)
• ci: show load tests inputs in the job summary (#50393)
• fix: preserve call hierarchy when modifying child process instances (#50366)
• feat: migrate SLACK_WEBHOOK_URL to Vault in Zeebe CI workflows (#50358)
• fix: rework message batch expiry to avoid scheduled command cache (#50346)
• fix: replace load test secret placeholder on redeploy (#50338)
• feat: replace BACKPORT_ACTION_PAT with Monorepo DevOps Automation GitHub App (#50333)
• When port is not available complete exceptionally, and log error. (#50280)
• feat: migrate AUTOMERGE_TOKEN to Monorepo Release GitHub App via Vault (#50266)
• feat: migrate SLACK_CAMUNDA_EX_CI_WEBHOOK from repo secret to Vault (#50257)
• feat: Update job runner (#50217)
• fix: use --reset-then-reuse-values in install-load-test Makefile targets (#50205)
• test: reenable contrast tests (#50171)
• fix: validate official docker images require explicit reuse-tag (#50162)
• fix: remove pretty-print JSON serialization in UserTaskCompletionVariableHandler (#50161)
• fix: silence log spam for user tasks without due date (#50153)
• Use same ratio for cache to total memory as in 8.7. (#50141)
• perf: batch RocksDB transactions in ExportersState.setExporterState (#50138)
• fix: add grace period to OAuth token validity and synchronize retry refresh (#50124)
• feat: add disabledQueries config to filter out read benchmark queries (#50111)
• feat: configure minimumReleaseAge for renovate (#50101)
• test: remove axios dependency (#50100)
• fix: use conservative memory tracking for exporter flush (#50054)
• fix: handle non-numeric formKey in FlowNodeInstanceMetadataBuilder (#50017)
• fix: track RPI usage metrics for process instances started via message, timer, and signal (#50012)
• feat: add trace logging for entity size in ES bulk request payloads (#49964)
• fix: handle assignment/unassignment/completion notification for lack of permissions (#49872)
• ci: split GH release into 2 parts + create changelog file + fisrt update assests (#49768)
• fix: remove all registered meters when ...

version:8.7.27

Bug Fixes

Misc

• Scroll contexts leaked in ElasticsearchUtil.scrollWith on error paths (#50892)
• Scrolling gets stuck in Tasklist v1 with large process list and search returns 500 error (#44583)
• Stop the OC when port 26502 is already in use (#42581)
• Operate (SM OpenSearch) does not show link to executed decision instance (#41929)
• The C8Run Docker distribution can't be shut down using './shutdown.sh --docker' command (#37323)
• Scheduled commands cache uses the wrong intent for caching message expiry (#15574)

Maintenance

• CI: move all repository secrets to Vault (#18211)

Task

• Passing correct tags for components in release load test (#50323)
• Schedule load tests in a single availability zone (#50028)
• Job throw error and cancel processors use TypedRecordProcessor (#40924)

Merged Pull Requests

• refactor: replace use-official-docker-images with orchestration-tag and add per-component image tags (#50417)
• ci: show load tests inputs in the job summary (#50393)
• feat: migrate SLACK_WEBHOOK_URL to Vault in Zeebe CI workflows (#50358)
• fix: rework message batch expiry to avoid scheduled command cache (#50346)
• fix: replace load test secret placeholder on redeploy (#50338)
• feat: replace BACKPORT_ACTION_PAT with Monorepo DevOps Automation GitHub App (#50333)
• When port is not available complete exceptionally, and log error. (#50280)
• feat: migrate AUTOMERGE_TOKEN to Monorepo Release GitHub App via Vault (#50266)
• feat: Update job runner (#50217)
• fix: use --reset-then-reuse-values in install-load-test Makefile targets (#50205)
• fix: validate official docker images require explicit reuse-tag (#50162)
• perf: batch RocksDB transactions in ExportersState.setExporterState (#50138)
• deps: Downgrade surefire plugin version from 3.5.5 to 3.5.2 (#50098)
• test: assert global state restore in backup-restore test (#49837)
• fix: Loadtest Container Names (#49568)
• ci: pin load tests to a single availability zone (#49541)
• feat: Add message sending to C8 cluster (#47164)
• fix: C8Run Docker shutdown docker fix (#41403)
• refactor: replace usages of CommandProcessor with TypedRecordProcessor (#41220)
• Refactor CommandProcessor-implementing Job processors (#34970)

What's Changed

• [Backport stable/optimize-8.7] fix: bootstrap full reactor to resolve unpublished SNAPSHOTs by @backport-action in #50277
• [backport 8.7] build: remove obsolete workflow for publishing C4 bundle to npm by @OmranAbazid in #50564
• deps(optimize): bump httpcore5 5.3.4 → 5.3.6 and netty 4.1.130 → 4.1.132 by @mustafadagher in #50777
• build: mergeback 8.7.19 to stable by @konstantinos-mylosis in #50503

Full Changelog: 8.7.19-optimize...8.7.20-optimize

version:8.9.0

Enhancements

Java Client

• API: Allow filtering and sorting by BusinessId in process instances search queries (#47246)
• Support handling GET requests with empty responses in the java client (#37243)
• Use OffsetDateTime for date-time fields in java client (#33678)

Misc

• feat: Add tests for multi-JWKS support (#47217)
• feat: Support multiple JWKS endpoints per OIDC issuer (#47213)
• Use the businessIdUniqueness config to toggle validation logic in the engine (#45715)
• Expose configuration property for business Id uniqueness validation (#45714)
• Engine: Propagate Business Id to child process instances (Call Activity: inherit-only for MVP) (#44980)
• Gateway: Stable partition routing for Business Id uniqueness (#44979)
• API: Support Business Id on process instance start (REST + gRPC) (#44977)
• Optimize: support extraConfiguration via SPRING_CONFIG_IMPORT (parity with orchestration) (#44820)
• [Spring SDK] Workflow for compatibility tests (#41948)
• [Spring SDK] Add integration tests (#41947)
• [Spring SDK] Configure revapi (#41941)
• [CPT] Compatibility tests workflow (#41917)
• [Camunda Client] Workflow for compatibility tests (#41916)
• [Camunda Client] Adjust acceptance test pom with new testing profile and client versioning (#41914)
• [Camunda Client] Create new annotation @CompatibilityTest for testing (#41913)
• [CPT: DSL] Assert process instance (#41298)
• [CPT: DSL] Throw BPMN error (#41267)
• [CPT: DSL] Create process instance (#41230)
• Improve backup/restore usability: always ensure missing index templates are created (>=8.8) (#40675)
• Restore to match RDBMS (#40233)
• [EPIC]: As a engineer I want to run a load test for a specific supported version (#38829)
• Mark batchOperations as FAILED if all partitions have failed (#37821)
• Introduce a write latency metric for the meta store (#36348)
• Allow users to always open variable value editor in a completed process instance on Operate (#32516)
• Handle user experience for "no permission to operation" (#29895)
• [CPT] I can use a test environment globally (#29131)
• Use element instances search C8 API endpoint for instance history in Operate UI (#27330)
• [CPT] Assert that a message was correlated to a process instance (#23291)
• [CPT] I can add a custom containers to the runtime (#23158)
• [CPT] Generate a process coverage diagram (#23153)
• I can download the diagram as bpmn xml file (#22005)
• Cluster health endpoint (#20921)
• [CPT] I can deploy resources via test annotation (#19178)

Bug Fixes

Broker

• Deploying too many versions of a large process can lead to startup problems (#11253)
• RoleChangeListener is not called after becoming Leader (#4409)

Gateway

• NullPointerException in MCP gateway for null userTaskKey in getUserTask (#48201)
• NullPointerException in MCP gateway for null userTaskKey in searchUserTaskVariables (#48197)
• NullPointerException in MCP gateway for null incidentKey in resolveIncident (#48196)
• Partition retry inconsistency between gRPC and REST gateways for process instance creation (#46756)
• Gateway can easily overwhelm broker with handling blocking jobs (long polling) (#7659)

Java Client

• [CPT] Slow test execution and test failures if the client prefers REST (#45667)
• Client failure handling ends in Stackoverflow (#34597)

Misc

• SUSPEND audit log entityKey does not match batchOperationKey (#50311)
• Evaluate an expression API response schema expects result to be unknown, not null (#50091)
• Scheduled flush timer in CamundaExporter get starved under sustained load (#49820)
• Failure to deserialise DbCheckpointState (#49812)
• NullPointerException for OperationType when handling BatchOperationChunkRecordValue (#49765)
• Task is not highlighted after navigating to History tab (#49707)
• Audit Log API: Sentinel Values Returned Instead of null (#49564)
• RESUME audit log entityKey does not match batchOperationKey (#49513)
• Inconsistent display: Identity name shows as 'Identity' in dashboard/app switcher but as 'Admin' in cluster details (Helm Chart, 8.9 snapshot) (#49512)
• Missing Navigation to Root Decision Instance in DMN Incidents (#49425)
• Batch Operation Details page cannot be scrolled (#49410)
• Delete decision instance returns 404 instead of 403 (#49280)
• Incident tab context remains visible after incident resolution in Operate details view (#49220)
• fix: POST /groups/search 500 response uses bare $ref to schema instead of proper Response Object (#49197)
• OpenAPI spec: DecisionEvaluationById.variables has incorrect description ("message variables") (#49190)
• Swagger UI POST requests results in a 401 with CSRF validation error (#48902)
• Delete decision instance API cannot use key returned by Get decision instance (#48867)
• Authorization search endpoint does not return page.startCursor and page.endCursor (#48857)
• USER_TASK authorization cannot be re-created (#48847)
• Fullscreen toggle does not work in Operate diagram controls (Chrome) (#48823)
• Operate: Navigating to parent process from Call Activity link causes page to continuously jump (#48807)
• UI: No margin in Input/Output Mappings panel list (#48801)
• Process instance details page doesn't handle permission errors on process instance history (#48728)
• MCP gateway: invalid search filter values cause 500 with Java class name leak (#48578)
• Recoverable exceptions in a ZeebeTransaction are not logged (#48538)
• Batch move of regular ad-hoc subprocess corrupts the process instance (#48530)
• Operate UI shows incorrect token count when moving MI instances to an outside element (#48527)
• Process Instance with tag does not complete (#48502)
• duplicate WHERE clause in MessageSubscriptionMapper getProcessDefinitionStatistics query (#48387)
• Tooltip not displaying correctly when hover over on the 3 dots for batch operation details page (#48386)
• [CPT] An assertion error in CamundaProcess...