Although the majority of mutter's build-time test and autopkgtest
failures appear to have been resolved during the 48.x series, looking
through build and autopkgtest logs there is one failure mode that has
not been accounted for.

In one autopkgtest on armel,
https://ci.debian.net/packages/m/mutter/testing/armel/61424275/ we see
this failure:
...

For further details of that failure we have to look in the
$AUTOPKGTEST_ARTIFACTS (artifacts.tar.gz on ci.debian.net) and read
artifacts/mutter-16_surface-scale-tests.test.txt, quoted in full here:

An equivalent failure was seen for 48.4-1 in experimental on
reproducible-builds.org, although only for the second of the two builds
that r-b does (in the first build, all tests passed):

This has not been seen on official buildds, and has not been seen on
ci.debian.net in the dozens of autopkgtest runs that were done since
May, so my suspicion is that it's a race condition or other
timing-related issue that rarely happens. Perhaps there is a race
between the mutter compositor setting the cursor scale to its correct
nonzero value, and the client retrieving it?

Assistance in reproducing or debugging this test failure would be
appreciated.

    smcv