fix(cloudflare): harden canonical RSC warmup end to end

[james-elicx]

Summary

Consolidates the canonical warmup hardening previously split across #3021–#3040. This range ends at the first clean, self-contained deployed boundary: earlier heads could begin filling cache keys while version-override routing was still mixed, which is fixed by the staged-readiness work included here.

Response and build validation

• validate effective Cloudflare cache policy, shared freshness, cache admission status, cookies, content type, exact supported Vary, and immutable build identity
• expose generic adapter-owned build identity at the outer App and Pages response boundary
• retry only failed staged requests after the initial queue and distinguish transient old-build responses from permanent same-build failures
• keep strict and best-effort warming bound to the uploaded Worker build

Canonical request ownership

• align concrete App/Pages and hybrid-i18n ownership between browser navigation, runtime routing, and warm discovery
• keep rewrite-affected, parallel/interception-only, and other contextual requests hashed rather than collapsing them into canonical entries
• prove Link, router.prefetch(), router.push(), explicit full prefetch, and soft navigation reuse on a real deployed Worker
• isolate each deployed proof attempt behind a fresh Worker/cache namespace

Discovery and staged deployment

• discover child routes using parent-generated params and exclude rewrite/redirect-owned public paths
• discover localized Pages keys with basePath and trailingSlash applied correctly
• warm the concrete production route/custom domain reported by Wrangler without changing TPR parsing
• derive version identity from upload output and fail closed when deployment state is unavailable
• establish stable uploaded-version readiness with unique no-cache probes before sending any canonical cache-fill request
• enforce strict/no-promote staging invariants, configured retry budgets, and complete fail-closed discovery

The three findings raised on #3038 are resolved by #3039 and #3040 within this consolidated range.

Scope

Only canonical RSC/ISR CDN warm request identity, build-time warm discovery, response validation, and Cloudflare staged deployment. No middleware, TPR, development-mode, cross-version-cache, local page rendering, or generic caching changes.

Validation

The retained commits keep their focused client, ownership, cache, adapter, discovery, deploy, version, build, and deployed Playwright coverage. This exact head passed all 67 checks, including strict Worker prewarming and browser reuse.

Consolidates #3021 through #3040. Stacked on #3020.

[github-actions]

Example	Preview	Production	Original
app-router-cloudflare	preview	production
pages-router-cloudflare	preview	production
app-router-playground	preview	production	original
realworld-api-rest	preview	production
nextra-docs-template	preview	production
benchmarks	preview	production
hackernews	preview	production	original
workers-cache	preview	production
web	preview	production

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/@vinext/cloudflare@3040

npm i https://pkg.pr.new/create-vinext-app@3040

npm i https://pkg.pr.new/@vinext/types@3040

npm i https://pkg.pr.new/vinext@3040

commit: 22c8e57

[github-actions]

Performance benchmarks

Compared 22c8e57 against base 6162f02 using alternating same-runner rounds. Next.js was unchanged and skipped.

1 improved · 0 regressed · 5 within ±1.5%

Scenario	Framework	Baseline	Current	Change
Client bundle size (gzip)	vinext	142.2 KB	142.2 KB	⚫ +0.0%
Client entry size (gzip)	vinext	129.5 KB	129.5 KB	⚫ +0.0%
Dev server cold start	vinext	2.22 s	2.18 s	🟢 -1.8%
Production build time	vinext	2.31 s	2.34 s	⚫ +1.4%
RSC entry closure size (gzip)	vinext	116.0 KB	116.0 KB	⚫ -0.0%
Server bundle size (gzip)	vinext	197.5 KB	197.5 KB	⚫ -0.0%

View detailed results and traces

_{🟢 improvement · 🔴 regression · ⚫ change below 1.5% · paired base/head}