[BUG]: saving state stalls the orchestration loop each iteration #226

Activity

1. gjanco

   added a parent issue on Jun 11, 2026

   • [EPIC] Performance #206
2. Shearerbeard commented on Jul 17, 2026

   

   Shearerbeard

   on Jul 17, 2026

   Collaborator

   More actions

   Vetted 2026-07-17 against an instrumented main build. The drain await is real (run_iteration blocks on persistence.drain once per iteration, orchestrator.rs:3724) but measured cost was 0ms in all nine sampled iterations. The fast path returns immediately when no writes are in flight, and the worst case is capped by persistence_drain_timeout_ms (default 2000ms). The 50-200ms estimate has no support in practice. Closing. If large-artifact workloads ever show drain time in a profile, #218's instrumentation is the way to see it.
3. Shearerbeard

   closed this as not plannedon Jul 17, 2026
4. Shearerbeard

   changed the title [-]Persistence drain barrier[/-] [+][BUG]: saving state stalls the orchestration loop each iteration[/+] on Jul 17, 2026