What does this PR do?

Stabilizes the shared test failures that remained after #21012 landed on main.

The fixes are intentionally small and CI-focused:

• preserve background-review agent messages before closing the review agent, so the self-improvement summary can still report actions
• reset auxiliary-client unhealthy-provider cache between tests, matching the existing runtime-main reset
• isolate hermes update tests from developer-local lazy backend activation state
• update the provider discovery count for the newly registered provider profile
• make compression-feasibility tests independent from local provider config and custom-provider attrs
• rehydrate persisted context summaries even when the handoff sits in the protected head after resume

Related Issue

Related to the CI gate for the hermes-lcm context-engine merge train, including stephenschoettler/hermes-lcm#133. This does not close that issue.

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

• run_agent.py: capture review_agent._session_messages before closing and clearing review_agent.
• agent/context_compressor.py: find persisted handoff summaries from the first non-system message through the compression window, not only from the computed compress-start boundary.
• tests/conftest.py: reset agent.auxiliary_client unhealthy-provider state per test.
• tests/hermes_cli/test_update_autostash.py: no-op active lazy-backend refresh in update-autostash tests.
• tests/providers/test_plugin_discovery.py: update the provider registry assertion from 33 profiles to 34.
• tests/run_agent/test_compression_feasibility.py: mock auxiliary provider resolution and include custom_providers in expected calls.
• tests/agent/test_context_compressor_summary_continuity.py: keep a regression shape that actually compresses new turns after a persisted handoff.

How to Test

1. Run the focused current-main failing set:
   ./scripts/run_tests.sh -n 0 tests/agent/test_auxiliary_client.py tests/agent/test_context_compressor.py tests/agent/test_context_compressor_summary_continuity.py tests/hermes_cli/test_update_autostash.py tests/run_agent/test_provider_parity.py tests/providers/test_plugin_discovery.py tests/run_agent/test_background_review.py tests/run_agent/test_compression_feasibility.py -q --tb=short
2. Run file sanity checks:
   git diff --check && python -m compileall -q agent/context_compressor.py run_agent.py tests/conftest.py tests/agent/test_context_compressor_summary_continuity.py tests/hermes_cli/test_update_autostash.py tests/providers/test_plugin_discovery.py tests/run_agent/test_compression_feasibility.py
3. Run lint on touched files:
   python -m ruff check agent/context_compressor.py run_agent.py tests/conftest.py tests/agent/test_context_compressor_summary_continuity.py tests/hermes_cli/test_update_autostash.py tests/providers/test_plugin_discovery.py tests/run_agent/test_compression_feasibility.py

Validation Status

• Focused failing set passed: 378 passed in 62.78s.
• git diff --check passed.
• python -m compileall -q ... passed for touched files.
• python -m ruff check ... passed for touched files.
• Full pytest tests/ -q is not claimed here. A broader local non-integration run on this branch still showed unrelated current-main failures outside this focused unblocker scope, so this PR keeps the validation claim narrow.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: Arch Linux, Python 3.11 venv via scripts/run_tests.sh

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — N/A
• I've updated cli-config.yaml.example if I added/changed config keys — N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — no platform-specific code paths added
• I've updated tool descriptions/schemas if I changed tool behavior — N/A

Screenshots / Logs

Focused validation output:

378 passed in 62.78s (0:01:02)
All checks passed!