Croct fetch in my loader adds latency when the network is slow, how do I cap it?

There is a timeout for exactly this. The JS SDK default is 5000ms, which is why a bad upstream can sit there a while before failing. You can bring that down globally with defaultFetchTimeout on the plug options, or pass a timeout per fetch when one slot has a tighter budget than the rest:

const {content} = await context.croct    .fetchContent('home-hero@1', {timeout: 500});

Set it to the latency budget you actually have for that route and the loader stops waiting past it instead of dragging TTFB with the upstream.

The timeout on its own only bounds the wait, it does not decide what renders when the wait runs out. Pair it with a fallback so a slow or failed fetch still serves default content rather than erroring:

const {content} = await context.croct    .fetchContent('home-hero@1', {timeout: 500, fallback: defaultHero});

Since resolution is server-side in the loader, that fallback renders straight into the HTML with no client-side flicker. In the normal case none of this even triggers, the documented end-to-end response times are under 90 milliseconds at P95, so the timeout is really just the floor under your worst case.