@smcv these are not marked flaky by upstream, but by us via https://salsa.debian.org/gnome-team/mutter/-/blob/debian/master/debian/patches/debian/tests-Tag-unstable-tests-as-flaky.patch so while those are indeed flaky in some scenarios, upstream also thinks that they should pass, so I was thinking that it's actually better to do retries than ignore possible real failures.

Hmm, OK. I can confirm that they can fail even on amd64 when under heavy load (building two packages involved in the GNOME 3.38 transition in parallel VMs, in parallel with importing a new mozjs version into git) so the failure is probably timing-related.

Yeah, I also got one or two cases where it happened... But I notified the upstream developer about, but he couldn't reproduce.

In any case that's the reason why I removed amd64 from the list as I think that is preferable to monitor such failures more than ignoring them forever (while pressuring upstream) :)