test(cache): add App Router Response Store load harness

[james-elicx]

Scenario

This adds an isolated vinext App Router load-test app for the production Workers Response Store path. It intentionally uses unique route IDs so each lifecycle starts from a cold key rather than measuring an already-warm Workers Cache hit.

Cacheable lifecycle

• Requests /cacheable/:id, an App Router page forced to render at runtime.
• next.config.mjs applies Cache-Control: public, s-maxage=15, stale-while-revalidate=15 only to this route.
• Verifies MISS on the first request, then the same render token on a stored HIT.
• Waits for freshness to expire, verifies stale UPDATING with the old token, then polls until a recovered HIT has a new token.
• This exercises cold metadata/body admission, Workers Cache population, SWR claim/regeneration, publication, and return to a hit.

Dynamic lifecycle

• Requests /dynamic/:id twice at the same URL.
• The page remains force-dynamic with no public cache-policy override.
• Verifies both responses are no-store, neither reports HIT/UPDATING, and each render has a different token.
• This exercises repeated Response Store misses that must never be admitted.

Load shape and output

• Defaults to 5,000 cacheable and 5,000 dynamic lifecycles.
• Uses a closed-loop concurrency ladder of 50, 100, 200, 400, and 800; this is not a fixed-arrival-rate RPS benchmark.
• Stops a scenario after its cumulative lifecycle failure rate exceeds 1%.
• Emits ignored, chart-ready JSON with per-second RPS, p50/p95/p99/max latency, statuses, cache states, errors, lifecycle totals, and admission/SWR transition timings.
• Enables logs, invocation logs, traces, and 1% head sampling on both the app and cache Workers.
• Keeps the manually provisioned Workers, Durable Object namespace, and R2 bucket after runs.

The in-directory README documents prerequisites, one-time bucket creation, build/deploy ordering, smoke and high-load commands, every runner option, JSON fields, and CSV export for charting.

Observed App Router experiment

• Cacheable lifecycles were clean through concurrency 200.
• At concurrency 400, 107 of 1,333 cacheable lifecycles failed (8.03%).
• A dedicated concurrency-800 cacheable run failed 871 of 1,000 lifecycles (87.1%).
• Dynamic lifecycles reached concurrency 800 with 37 of 1,668 failures at that level (2.22%).
• Failures included Durable Object is overloaded. Requests queued for too long.
• The dashboard showed a 249.42 Durable Object invocations/s minute-average peak. This is experiment context, not a precise single-instance capacity limit.

The generated experiment JSON and graph artifacts are not committed.

Validation

• node --check examples/response-store-demo/load-test/load.mjs
• App Router production build completed through all five vinext build stages.
• Live smoke: 1 cacheable and 1 dynamic lifecycle passed, 9 requests total, zero failures.
• git diff --check

[pkg-pr-new]

Open in StackBlitz

@vinext/cloudflare

npm i https://pkg.pr.new/@vinext/cloudflare@3299

create-vinext-app

npm i https://pkg.pr.new/create-vinext-app@3299

@vinext/types

npm i https://pkg.pr.new/@vinext/types@3299

vinext

npm i https://pkg.pr.new/vinext@3299

@cloudflare/workers-response-store

npm i https://pkg.pr.new/@cloudflare/workers-response-store@3299

commit: 2ded4f2

[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
response-store-demo	preview	production
static-export	preview	production
web	preview	production

[github-actions]

Performance benchmarks

Compared 2ded4f2 against base 3aa24cf using alternating same-runner rounds. Next.js was unchanged and skipped.

0 improved · 0 regressed · 6 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	129.6 KB	129.6 KB	⚫ +0.0%
Dev server cold start	vinext	3.35 s	3.32 s	⚫ -0.8%
Production build time	vinext	3.86 s	3.85 s	⚫ -0.2%
RSC entry closure size (gzip)	vinext	121.0 KB	121.0 KB	⚫ +0.0%
Server bundle size (gzip)	vinext	211.0 KB	211.0 KB	⚫ +0.0%

View detailed results and traces

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

[james-elicx]

Closing because this load experiment was not requested as a pull request. The harness and result work will continue locally.