PR 30 · commit 1265acf · report 2026-06-05T03:50:01Z

Preview wins PSI and payload. HAR gate passes with RSC warning.

Fresh HAR captured for deploy preview 30 and production. PageSpeed values parsed from supplied mobile reports. The preview cuts JS/image bytes and improves Lighthouse mobile score/LCP/TBT versus production, but fresh HAR still shows RSC over the 40s target and a long polluted session span from deploy-preview tooling/monitoring requests.

Gate statusPASSRSC warn
Gate RSC48.36starget ≤40s; stop >60s
PSI score Δ+2667 vs 41
HAR bytes Δ−6.24MBpreview vs prod

PageSpeed mobile

metricprodpreviewdelta
Performance4167+26
FCP1.5s2.4s+0.9s
LCP7.5s3.3s−4.2s
TBT1,570ms820ms−750ms
CLS0.0010−0.001
Speed Index9.3s4.4s−4.9s
Payload total5,759KiB3,044KiB−2,715KiB
Main-thread work8.2s5.3s−2.9s
Long tasks2011−9

Reports: preview · production

Fresh HAR: prod vs preview

metricprodpreviewdelta
Requests128126−2
Session span27.26s72.73s+45.47s
Body bytes14.45MB8.21MB−6.24MB
Summed request time190.91s119.04s−71.87s
Summed wait102.34s110.21s+7.86s
RSC requests1723+6
RSC bytes1.45MB746.7KB−~724KB
RSC summed time47.43s48.36s+0.93s
JS bytes8.11MB5.04MB−3.07MB
Image bytes2.64MB201.3KB−2.44MB

Gate run

PASS Batch 31 gate cleared.

  • Preview reachable: HTTP 200.
  • Header errors: 0.
  • Private-route leaks: 0.
  • /hedge/* duplicates >2: 0.
  • RSC summed time: 48.36s; warning only, under 60s hard stop.

Readout

Claim safe: preview is materially lighter and PageSpeed-mobile better. Claim not safe: all timing improved. HAR has RSC over target and slow duplicated document/RSC/monitoring requests.

Next work: reduce RSC prefetch/fan-out for /about and /world-cup; keep monitoring noise out of perf captures; retest preview without Netlify deploy-preview UI pollution.

Artifacts

Batch 31 gate cascade · Fresh prod-vs-preview cascade · Preview HAR · Prod HAR

Generated with scripts/batch-gate.ts and scripts/analyze-har-cascade.ts. Typecheck note from external checker used pnpm; project package manager is Bun.