For rust-gst-plugin-gtk4, debian/debcargo.toml has

[packages."lib+gst-gl"]
test_is_broken = true
[packages."lib+dmabuf"]
test_is_broken = true

(Let's ignore dmabuf since that will be removed once Debian is unfrozen).

I would expect only the --features gst-gl test to be marked as flaky.
The all-features test is marked flaky but see
https://bugs.debian.org/985762

I see in Cargo.toml that several features depend on gst-gl so I guess
that's why they were included as flaky. However, those feature
autopkgtests do pass. gst-gl is the only autopkgtest that fails. I
reported that specific failure upstream as
https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/issues/698

Similar to the other Debian bug, I can see why debcargo might want to
do this by default but I think it would be nice for a package
maintainer to be able to override that behavior to allow others tests
to avoid being marked as flaky.

Thank you,
Jeremy Bícha

this is intentional and documented, see debcargo.toml.example:

https://salsa.debian.org/rust-team/debcargo/-/blob/master/debcargo.toml.example?ref_type=heads#L233

# The effect is transitive to its reverse-dependencies, so e.g. if you specify
# this for feature A, and feature B depends on feature A, then feature B also
# implicitly has this set. To unset it on feature B (and its transitive rdeps),
# explicit set this to false for feature B as well.

you can override it, but if there's a lot of features depending on the
broken feature it's a bit verbose ;)