Verdict: RETIRED 2026-05-16. 0 wins / 589 losses across a 9-triangle hardcoded slot set — no resuscitation. The grinder ran long enough to accumulate 589 on-chain slot calls; every single one came back as a miss (slot mechanism succeeded, all 9 strategies returned 0 profit, gas burned). Lifetime net −4.40 WPLS is the sum of gas across 589 misses × ~$0.0075 each. No fires. No wins. No close-to-wins.

Failure mode #1 — the 9-triangle hardcoded set is a 2023 fossil. A static list of 9 PLSX-V1 triangles was where the alpha lived when PulseX was thinly arb'd. Three years in, every meaningful triangle on those 9 routes is hit deterministically by competitor bots in the same block we'd quote from. Our slot-call timing arrives after the spread is gone.

Failure mode #2 — even a successful slot call returned 0 in 589/589 attempts. This is the load-bearing fact. The slot mechanism (9 strategies fanned out, profit picker fires the best) found ZERO non-zero strategies across 589 attempts. That's not a tuning problem; the triangles themselves have no residual edge at our latency profile. No parameter change moves 0/589 to any-positive sample size.

Failure mode #3 — already abandoned operationally before this verdict. Last trade log entry: 2026-05-04. Last driver log: 2026-04-22. tmux session not running for weeks. The retirement decision was made implicitly by neglect; this conclusion is the formal record.

Fixable? No. Retired. Do not redeploy. If the slot-arb premise is ever revisited, it requires (a) a triangle set discovered dynamically from current PLSX-V1/V2/9inch pool surface, not the 2023 hardcoded set, and (b) private-mempool inclusion to beat the deterministic competitors. Neither is a tweak; both are different bots. Free the slot.