fix(cli): keep the DE parallel-router trial on until a real failure, not first engagement

Only consuming telemetry from one data point per install badly undersampled
the "routed" (successful) outcome — the far more common case. Changed
maybeConsumeDeParallelRouterTrial to only turn the trial off when the
router's OWN safety net actually fired (deParallelRouter === "reverted"),
not on a clean "routed" success. This runs the experiment on every eligible
render for an install indefinitely until it hits one real failure, then
stops for that install going forward — trading a slightly higher per-install
ceiling on experimental-path exposure for dramatically more successful-
routing telemetry volume across the fleet.

Also fixed a related edge case while updating this: a render that merely
"routed" (router fired, self-verify never even tripped) but then crashed
for an unrelated reason (e.g. cancellation) no longer counts as a router
failure — only "reverted" (the router's fallback path actually engaged)
does. Cancelling a render isn't evidence the router is unsafe.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This commit is contained in:
Vance Ingalls
2026-07-09 21:49:21 -07:00
co-authored by Claude Sonnet 5
parent 37b6a4e7e5
commit 19f90b0b92
3 changed files with 67 additions and 29 deletions
+8 -6
View File
@@ -65,12 +65,14 @@ export interface HyperframesConfig {
skillsMissingCount?: number;
/**
* True once the DE parallel-router experiment ("HF_DE_PARALLEL_ROUTER")
* has actually engaged (routed or reverted — either produces telemetry)
* on a render from this install. The CLI enables the experiment for free
* on renders from a fresh install until this fires once, then never
* touches it again — a one-shot trial to get real-traffic router
* telemetry without requiring anyone to manually opt in via env var.
* See `renderLocal`'s `maybeEnableDeParallelRouterTrial`.
* has actually FAILED (its self-verify/generic-failure safety net fired —
* "reverted", not merely "routed") on a render from this install. The CLI
* enables the experiment for free on EVERY eligible render from a fresh
* install — not just once — to maximize real-traffic router telemetry
* (mostly successful "routed" outcomes) without requiring anyone to
* manually opt in via env var; only a real failure turns it off, and only
* for this install going forward. See `renderLocal`'s
* `maybeEnableDeParallelRouterTrial`/`maybeConsumeDeParallelRouterTrial`.
*/
deParallelRouterTrialFired?: boolean;
}