Friction dashboard: scanner 404s at warn tier were 45% of user friction
Measured 2026-08-10 off vUserFrictionEvents: navigation/page_not_found = 23 events, 45% of all 7d user friction. 22 of 23 unauthenticated, all classifiedReason=genuine-bad-route (classifier ran and called them genuine, so this is a wrong-call gap not a never-reached gap). Dashboard header promises 'Excluye trafico de rutina y sondeos de bots' while bot probes top it, burying the ONE authenticated hit at 4%. Fix: SCANNER_PROBE_RE block S adds .gitconfig/.git-credentials, next build-manifest, graphql (root/api/v1), rclone.conf, Vite @fs traversal. Held out deliberately: /completar-perfil (real broken link, PLUTO-655), /blog (app-plausible word), a random-slug debug-trigger path (provenance unestablished, may be a canary), and the single authenticated /mi-<pasted-external-url> hit.
Questions
Activity
-
CORRECTION (coder-pluto-cc + db-pluto-cc measurement, 2026-08-10): '/completar-perfil (real broken link, PLUTO-655)' in this WI's own description is FALSE — never-true premise, inherited by this WI and the PLUTO-655 test fixture independently, so they falsely corroborated each other. Path has never existed in Pluto (grep+git-log confirm) and has exactly 2 hits ever, 26 days apart, both facebookexternalhit/Twitterbot UA, no referrer, unauthenticated -- link-preview crawler traffic, not a human or a broken link. The hold-out (leave it pageable) still stands; only the 'real broken link' characterization is retracted.
-
Shipped 54019d1/v2.22.79, audit PASS:54019d1. PTD: deploy READY, prod, live version 2.22.79 (dpl_5uznAg4rSNUrRK3d4Cs4S3Ni86TR). SCANNER_PROBE_RE block S covers .gitconfig/.git-credentials, next build-manifest, graphql root/api/v1, rclone.conf, Vite @fs traversal. Correction note added re: /completar-perfil never-true premise.