See bug 1715503
Note, the test does fail anyhow in many places
https://searchfox.org/mozilla-central/source/testing/web-platform/meta/fetch/metadata/prefetch.https.sub.html.ini

This is not seen as failing in latest try run https://treeherder.mozilla.org/jobs?repo=try&revision=76f19083ee9a7d7d7f38f568e2292c5ef30b74f1 . Peter is enabling bfcache in parent today (Bug 1715300). We'll confirm and close tomorrow if this doesn't fail anymore.

This doesn't need to be M7a. The test is failing always everywhere, but in different way with bfcache.
It has been disabled.

Fission MVP might be even enough for this, or Future.

This appears to be realted to the cache headers being set by the WPTs in record-header.py here.

I ran the tests with and without fission using:

 MOZ_LOG="nsPrefetch:5" ./mach web-platform-tests testing/web-platform/tests/fetch/metadata/prefetch.https.sub.html

and in both cases the following log can be observed:

 D/nsPrefetch document cannot be reused from cache; canceling prefetch

See here: https://searchfox.org/mozilla-central/source/uriloader/prefetch/nsPrefetchService.cpp#211-219

It appears that there is no issue with fission or the sec-fetch logic here but rather an issue with how we interpret the caching headers in the context of a prefetch.

It appears that there is no issue with fission or the sec-fetch logic here but rather an issue with how we interpret the caching headers in the context of a prefetch.

Based on this comment ^^, this is not a sec-fetch-*problem - un-assigning Niklas ...

Moving from Fission MVP to Future backlog.

Fission is annotated as intermittently timing out on these tests:

https://searchfox.org/mozilla-central/rev/73330bf7355c0aef844c41d0d7eed2848e53c82f/testing/web-platform/meta/fetch/metadata/prefetch.https.sub.html.ini#3,6,11,16

We should either figure out why Fission is intermittently timing out or remove the annotations.