Fix flaky JetStream publish test in clustered environments

[lalinsky]

Summary

Fixes a timing race condition in the "JetStream publish basic message" test that was causing intermittent CI failures with NoStreamResponse errors.

Root Cause

In NATS clustered environments, when a stream is created, there's a brief propagation window before all cluster nodes are aware of the new stream. If a client publishes immediately after stream creation to a node that hasn't received the metadata yet, it receives a 503 No Responders error.

Changes

• Added 100ms delay in jetstream_test.zig after stream creation to allow cluster metadata propagation
• Added TODO comments in jetstream.zig for implementing proper retry logic per ADR-22
• Referenced nats.go's approach (250ms backoff, 2 retries by default)

Test Plan

• Existing test should now pass reliably in CI
• Delay matches pattern used in other JetStream tests (push, nak, etc.)
• Full retry logic implementation tracked for future work

[coderabbitai]

📝 Walkthrough

Walkthrough

Adds TODO comments documenting planned retry logic for 503 NoResponders errors in clustered environments, referencing ADR-22. Introduces a 100ms test delay to allow stream metadata propagation across cluster nodes before message publishing. No logic or control flow modifications.

Changes

Cohort / File(s)	Summary
Documentation & Planning src/jetstream.zig	Added 6 lines of TODO comments outlining retry strategy for NoResponders errors in cluster scenarios, with ADR-22 reference
Test Timing Adjustment tests/jetstream_test.zig	Inserted 100ms sleep after stream creation in "JetStream publish basic message" test to allow metadata replication; includes TODO referencing ADR-22

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

The changes are minimal and involve documentation plus a straightforward timing adjustment. The main consideration is understanding the cluster propagation timing requirement, but there's no complex logic to trace through.

Possibly related PRs

• Basic JetStream API support #11: Introduces the JetStream implementation that these TODO comments and test adjustments directly address.

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately describes the main fix—adding a delay to resolve a flaky test caused by timing issues in clustered NATS environments.
Description check	✅ Passed	The description clearly explains the root cause (metadata propagation delay in clustered environments), the fix applied (100ms delay), and references TODO comments for future retry logic work.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

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

✨ Finishing touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch fix-jetstream-publish-race-condition

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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