> Comments on the principle of keeping build and test results clean?
Agreed. This was asserted as a responsibility of committers in the past,
though occasional carelessness has always occurred.
It should be noted that various NPTL tests have often been flaky on common
systems due to kernel bugs that are outside out control. So some tests
might only be expected to reliably pass on a very recent kernel. It's
probably the case that the last such fixes were long enough ago that "very
recent" is not today applicable in any absolute sense, but that issue is
sure to arise again.