[docs] Fix production deploy

[eps1lon]

Didn't need any code changes to fix it. I'm leaving the change to vercel in regardless to get a feel of how much slower exporting every locale is. Actual analysis of the issue below.

production deploys are currently failing for next (e.g. https://app.netlify.com/sites/material-ui/deploys/5fd31f0b071ca50007de7b01). Opening to notify that we're aware of the issue.

post-mortem

Broken when merging #23745. Shortly after it another PR was merged skipping the netlify deploy so https://app.netlify.com/sites/material-ui/deploys/5fd2552ee192c200082e8622 is the first one failing even though it's unrelated.

Figuring out:

1. why is it not failing on PRs even though it fails with the english locale
2. Netlify only issue?

Looks like a caching issue. Notice how on https://app.netlify.com/sites/material-ui/deploys/5fd31f0b071ca50007de7b01 we fail with a cache hit. However, https://app.netlify.com/sites/material-ui/deploys/5fd39c682427ad0c51e57510 is passing with a cache miss (I re-deployed with a cleared cache). The failing pages got introduced on a skipped build. This comes down to nextjs caching things like the page and routes manifest. However, the cache key for those is the yarn.lock file so we could have a cache hit even though pages changed.

I'm not sure whether netlify should never skip on next or if nextjs should re-generate those files all the time. If nextjs actually caches these manifests that aggressively then we should get build failures everytime we change pages. That should be a frequent problem but considering next export is an "advanced use case" it might be a rare bug for them.

[mui-pr-bot]

No bundle size changes

Generated by 🚫 dangerJS against ad384d9