[codex] Prewarm route transform workers

[ScriptedAlchemy]

Summary

• Keep route transform worker slots lazy so dev startup no longer creates the whole pool on the first transform executor use.
• Prewarm the full route transform worker pool from onAfterCreateCompiler in dev mode, before lazy route compilation requests hit the executor.
• Add coverage for lazy slot creation and explicit prewarm behavior.

Why

The first parallel transform implementation could defer web/node environment builds, but route transform workers still had a startup cliff. Pure lazy worker slots improved initial dev server readiness, but browsing into lazy-compiled routes could still pay worker creation cost. I tested full-pool and yielded/batched prewarm variants to find the best tradeoff.

Benchmark Notes

Large dev benchmark medians from local runs on this machine:

Variant	wall ms	ready ms	route total ms	wall p95	first route ms
old custom worker baseline	8570.3	5787.5	2576.0	9039.4	2301.3
lazy worker slots	7431.2	4957.3	2342.6	7742.1	2084.0
prewarm on onAfterCreateCompiler	7417.9	4872.7	2354.5	7670.0	2094.2
yielded prewarm	7509.4	5001.2	2335.3	7592.5	2071.7
prewarm on onBeforeDevCompile	7828.3	5009.6	2366.7	8138.8	2105.9
batched yielded prewarm	7711.4	5048.0	2463.8	8805.8	2193.4

onAfterCreateCompiler full-pool prewarm was the best overall choice: fastest median wall and ready time among the tried prewarm variants, with better p95 than lazy-only, while avoiding lazy route requests being the first thing to create worker slots.

Validation

• pnpm exec tsc --noEmit
• pnpm test tests/parallel-route-transforms.test.ts tests/index.test.ts tests/features.test.ts tests/build-output-transforms.test.ts
• pnpm build
• git diff --check

[pkg-pr-new]

Open in StackBlitz

pnpm add https://pkg.pr.new/rsbuild-plugin-react-router@f6f5a93

commit: f6f5a93

[github-actions]

Benchmark Results

Compared PR head 668fbf8 against base 6a49a94.

Total median wall time: 12.24s -> 12.10s (-1.1%, 1.01x speedup)
Compiler ready median: 10.04s -> 9.94s (-0.9%)
Route load median: 2.14s -> 2.08s (-2.9%)

Benchmark	Base total	Head total	Delta	Head ready	Head routes	Speedup	Head RSS p95
large-355-ssr-esm	12.24s	12.10s	-1.1%	9.94s	2.08s	1.01x	-

Profile: large; mode: dev; iterations: 10; warmup: 0.
Workflow run