This has historically failed on Debian autobuilders, and as a result we have had a patch for a long time that turned a failure here into a g_test_skip(). It's not clear whether this still happens, so I'm now assessing whether the patch can be dropped; but if the prlimit() call can fail for whatever reason, it would be useful for the error message to say what limit we were trying to set.
Part of one of my intermittent attempts to stop Debian having quite so many non-upstreamed patches.