Checkpoint sync state on all cancellation types

[arreyder]

Summary

• Sync loop now checkpoints in-memory progress on all cancellation types (e.g. worker shutdown SIGTERM), not just DeadlineExceeded (run duration timeout)
• Uses context.WithTimeout(context.Background(), 30s) for the checkpoint since the caller's context may already be cancelled — same pattern as c1file.Close()
• Without this fix, up to 10s of sync progress (minCheckpointInterval) is silently lost on every rolling deploy

Context

Combined with the c1-side WorkerStopTimeout fix (c1 PR #14204), this ensures the full graceful shutdown chain works:

1. Worker shutdown → Sync() checkpoints state before returning (this PR)
2. syncer.Close() → WAL checkpoint + local saveC1z
3. c1's manager.SaveC1Z() → uploads to S3 within the 25s grace period

Test plan

• New TestCancellationCheckpoints — cancels context mid-sync, verifies checkpoint persists, resumes successfully
• All existing pkg/sync/... tests pass

🤖 Generated with Claude Code

Summary by CodeRabbit

• Bug Fixes

  • Improved sync cancellation handling to save progress through checkpointing before exiting

• Tests

  • Added test coverage for cancellation and checkpoint resume workflows

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

---

Walkthrough

The sync package's cancellation handling now ensures checkpoints persist before exiting when context cancellation occurs. A new test validates that syncs can resume from checkpoints after mid-operation cancellation.

Changes

Cohort / File(s)	Summary
Cancellation Checkpoint Logic pkg/sync/syncer.go	Modified context cancellation handler to perform a checkpoint operation with a 30-second timeout using a background context before returning. Logs informational message on cancellation and errors on checkpoint failure, while still returning the original cancellation error.
Cancellation Resume Test pkg/sync/syncer_test.go	Added TestCancellationCheckpoints test with supporting cancelOnResourcesMockConnector mock type. Validates that syncs can be interrupted, checkpointed, and resumed from that checkpoint in subsequent sync invocations.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~22 minutes

Poem

🐰 When cancellation calls so loud,
We checkpoint with determined cloud,
A timeout guards our graceful flight,
Resume from where we paused mid-night,
Sync flows on, ever bright! ✨

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title directly matches the main objective: checkpointing sync state on cancellation types instead of only on timeout.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

• 📝 Generate docstrings (stacked PR)
• 📝 Generate docstrings (commit on current branch)

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch crr/fix-shutdown-c1z-race

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[arreyder]

Superseded by #738 which rewrites this fix against current main (the sync loop was refactored into sequentialSync/parallelSync since this branch was created).

[linear]

OPS-371 Checkpoint sync state for all cancellation types