# exec 10 Oct 2026: first look at fmmc5 execution (mirror data, 43 live rows, no code changed)

## 1. Stop fills below 0.5x
- Since 8 Oct 20:00 (29 stops with stop_slip): median fill 0.48x, median slip vs trigger value -0.8% (the stop itself works). 6 of 29 slip >10%, 8 of 29 fill <0.4x.
- Cause is time, not the order: STOP_PATH shows cross->sent 1.5-5.6 s (cross->seen 1.1-3.7 s is the biggest part: block feed round + 0.25 s poll), then 1-3 s to fill; a falling coin keeps dropping (sell_gap quote-vs-fill 0.14-0.30 on the bad ones). Two rows are rugs (0.004x, 0.008x), unsellable at 0.5x whatever we do.
- GMGN swap latency is NOT the problem (gmgnd 245-275 ms).
- Possible fixes, none applied: (a) shave cross->seen (dedicated per-block call for the open pool); (b) a higher stop (0.6x) is a strategy call for Ben/analysts, replay first. A position is open now, so no restart/edit this run.

## 2. Entry misses / 429
- The 19.5x (0x48b07961, bought 7 Oct 18:31) was CAUGHT: sold 19.51x. The 429s in mc5.log are Alchemy feed refusals (2 events, 8 and 9 Oct), no longer the source (feed is drpc, 0 failed calls). fmmc5 does no GMGN reads (quotes from chain); GMGN is only the swap POST via gmgnd. 2 SWAP_ERR total (HTTP 400 ValueLoss, 8 Oct).
- A shared GMGN gate on fmmc5 would only delay stop sells: NOT added. fml2's 429s came from GMGN quote calls, which fmmc5 does not make. The 8 Oct 19.5x miss is probably another coin (fml2?): need its address.
- late_scan 'read operation timed out' errors on 10 Oct 07:53-09:01 (5 pools): check they do not drop a qualifying coin.

## 3. Entry delay
- fill_delay median 3.4 s, max 12.0 s (2 of 43 above 10 s). Comparison against the entry rule still to do.