Fixes #xyz

Main Issue: #xyz

PIP: #xyz

Motivation

log: https://gist.github.com/Denovo1998/c48c58c1de0270a4008ae77b6161adfb

• Running the full ExtensibleLoadManagerImplTest occasionally fails at testDisableBroker. After calling disableBroker(), the broker unregisters and the channel cleans ownerships. However, subsequent checkOwnershipAsync/getOwnerAsync against the disabled broker could still trigger a liveness verification on metadata-based table view, which throws a MetadataStoreException (wrapped by ExecutionException), causing flaky failures in full-suite runs.
• Meanwhile, ServiceUnitStateChannelTest has explicit expectations on disabled/closed channels:
  • isOwner should still return true for Owned or Splitting states if the local broker is the designated owner/source,
  • getOwnerAsync should not complete immediately for Assigning (dst=local) or Releasing (dst=local), i.e., the future should remain unfinished to simulate “waiting for ownership”.
• We need to unify the disabled/closed-channel behavior to short-circuit ownership queries without metadata liveness checks, while preserving the above per-state semantics, eliminating flaky behaviors and aligning with test expectations.

Modifications

• Refined ServiceUnitStateChannelImpl#getActiveOwnerAsync to handle the disabled/closed channel states without performing liveness verification:
  • Owned, Splitting: return the owner (if any) as-is, so isOwner() on a disabled/closed channel remains true when appropriate.
  • Assigning, Releasing:
    • If the dst is the local broker, return an unfinished future (via dedupeGetOwnerRequest) to reflect “waiting for ownership”, matching testActiveGetOwner’s expectation (future is not done).
    • If the dst is another broker, return that broker ID to facilitate upper-layer redirection.
    • If no dst present, return Optional.empty().
  • Other states (Init, Free, Deleted, …): return Optional.empty().
• For non-disabled/closed channels, keep the original behavior unchanged:
  • If the broker registry is not registered:
    • On metadata-based table view, still fail with MetadataStoreException (unchanged).
    • On in-memory table view, return owner directly (unchanged).
  • Otherwise, continue to dedupe and validate liveness using registry.lookupAsync(newOwner) (unchanged).
• This change:
  • Prevents MetadataStoreException in disabled/closed scenarios, so testDisableBroker no longer fails randomly.
  • Keeps isOwner semantics on disabled/closed channels for Owned/Splitting.
  • Keeps Assigning/Releasing(dst=local) getOwnerAsync as an unfinished future, matching test expectations.
  • Does not affect the normal STARTED channel behavior nor the existing metadata-based/in-memory table view behaviors.

Verifying this change

• Make sure that the change passes the CI checks.

(Please pick either of the following options)

This change is a trivial rework / code cleanup without any test coverage.

(or)

This change is already covered by existing tests, such as (please describe tests).

(or)

This change added tests and can be verified as follows:

(example:)

• Added integration tests for end-to-end deployment with large payloads (10MB)
• Extended integration test for recovery after broker failure

Documentation

• doc
• doc-required
• doc-not-needed
• doc-complete

Matching PR in forked repository

PR in forked repository: Denovo1998#11