[ChromaBeast]

What changed

Follow-up to #2, whose Windows CI job failed after merge. Two commits:

1. test(metrics): require the disk-free gauge only where it is reported. metrics.DiskFreeBytes is unix-only (the gauge is omitted elsewhere, as documented), but TestMetricsAuthAndFormat required beastdb_disk_free_bytes unconditionally, so it failed on Windows CI with missing "beastdb_disk_free_bytes ". The test now checks the gauge only when DiskFreeBytes reports support. This one is a certain fix.
2. perf(engine): let the first writer lead its own commit group. Instead of waking a dedicated committer goroutine for every write, the writer that finds no commit in progress runs the queued group on its own goroutine; if more writers queued meanwhile, a drain goroutine serves them. Group-commit semantics are unchanged (one fsync per group, engine lock held across it, observer notified in LSN order, one group at a time so queue order is LSN order). Close now waits for the in-flight group and rejects later writes.

Why

TestEngineSystemWorkloadBenchmark also failed on Windows CI in #2: 398 ops/s against its fixed 1000 ops/s floor, with p95 latency 16.5 ms. Earlier Windows runs on main measured 1,351 and 1,923 ops/s with p95 about 3.5 ms. My hypothesis is that waking a separate goroutine for every uncontended write costs up to a timer tick (about 15.6 ms) on Windows, which matches the p95.

Things a reviewer should know

• The commit-path change is unproven. I cannot run Windows locally, and on macOS it did not change the local numbers. This PR's Windows CI run is the experiment. If Windows throughput does not recover, drop the second commit; it is separate for that reason. The throughput floor in the test is not changed.
• What I could and could not establish locally: on the original code (d8c1280) the benchmark gets about 1,560 to 1,710 ops/s in a full parallel go test ./... -race run; on the merged branch about 830 to 930 (fix branch included). Run alone, the two are within about 4 to 7% of each other. I counted fsyncs (one per write, same as before) and ruled out the beastbench tests and synthetic fsync load as the cause. The remaining difference under a full parallel run is unexplained.
• Tests pass with the race detector, including the group-commit crash-recovery, conditional-atomicity and write-after-close tests. The benchmark test still fails in a full parallel run on this laptop, as described above.
• Supersedes Fix Windows CI after #2: metrics test, single-writer commit path #3, which was based on commit hashes from before main was rewritten and conflicted. This branch is the same two commits rebased onto the current main.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-12T12:09:37.817635Z	f851bd6	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.