“It usually works… but the tail is brutal.”
Drop in a small JSONL receipt file — get a ranked fix order in ~5 minutes.
If retries are spiking, 429s are showing up, or you can’t tell which call is killing your p99 — Pitstop Scan shows where the pain concentrates and what to cap first.
Don’t have receipts yet?
Run make demo to generate a synthetic sample and see the output before you commit anything real.
Want the fix order mapped to your specific hazards?
Send output/pitstop_pack_agg.zip — derived summaries only, nothing raw —
and get back a 1-page Correctness Findings doc within 48 hours.
decision_event.v1 supported.