Rolldown server build stalls / never completes with high volume of SFC <style> blocks

[nicholaspufal]

Describe the bug

I am doing …

Running a production Nuxt 4 app build (yarn build) that uses Vite 8.0.0-beta.16 with Rolldown — the issue below is reproducible with npm:rolldown-vite@latest, no difference.

The app has SSR enabled and roughly 680 Vue SFCs, each with a non-trivial amount of component-scoped CSS in <style> blocks.

The same app built with standard Vite (Rollup) completes successfully.

What I expect is …

The build to finish: client and server bundles should be produced, even if it takes a while under that CSS load.

What actually is happening is …

With Rolldown, the server build appears to hang: it never completes, CPU drops to ~0%, and memory usage plateaus at around half of the memory available to the build process.

The process does not recover; it has to be killed (e.g. kill -9). If I remove all <style> blocks from those SFCs (keeping the same ~680 components and structure), the build completes with Rolldown, so the problem seems tied to Rolldown’s handling of a large amount of CSS from many Vue SFCs during the server build.

Reproduction

https://github.com/nicholaspufal/sample-nuxt-app-vite-rolldown-issue

Steps to reproduce

https://github.com/nicholaspufal/sample-nuxt-app-vite-rolldown-issue
☝️ Clone, yarn install, yarn build.

With the current styling in that app, the server build hangs; with styles stripped from those SFCs the build succeeds.

System Info

System:
    OS: macOS 26.3
    CPU: (12) arm64 Apple M2 Max
    Memory: 135.66 MB / 64.00 GB
    Shell: 5.2.15 - /opt/homebrew/bin/bash
  Binaries:
    Node: 24.13.0 - /Users/npufal/.nvm/versions/node/v24.13.0/bin/node
    Yarn: 4.9.2 - /Users/npufal/.nvm/versions/node/v24.13.0/bin/yarn
    npm: 11.6.2 - /Users/npufal/.nvm/versions/node/v24.13.0/bin/npm
    Deno: 2.3.3 - /opt/homebrew/bin/deno
  Browsers:
    Chrome: 145.0.7632.117
    Firefox: 136.0.2
    Safari: 26.3

Used Package Manager

yarn

Logs

output.log

Validations

• Follow our Code of Conduct
• Read the Contributing Guidelines.
• Read the docs and the Rolldown-related guide.
• Check that there isn't already an issue that reports the same bug to avoid creating a duplicate.
• Check that this is a concrete bug. For Q&A open a GitHub Discussion or join our Discord Chat Server.
• The provided reproduction is a minimal reproducible example of the bug.

[sapphi-red]

Would you try if setting ROLLDOWN_MAX_BLOCKING_THREADS=8 (or a bigger value) changes the behavior?

[nicholaspufal]

@sapphi-red Still stuck unfortunately.

I grabbed the max threads available via sysctl -n hw.ncpu (12 in my case), then ROLLDOWN_MAX_BLOCKING_THREADS=12 yarn build

[sapphi-red]

The thread number can be bigger than the number of your cpu threads, so would you also try a huge number like 256, 1024?

[nicholaspufal]

Tried 256, 512, 1024. Same.

When inspecting the process it's looking the same as without ROLLDOWN_MAX_BLOCKING_THREADS — CPU usage is at 0% and RSS just plateaus at ~850MB, while the build is stuck.