Just when you thought it was safe to go back in the water:
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=prairiedog&dt=2021-09-29%2022%3A05%3A44
which is complaining that the (misspelled, BTW) log message
'sucessfully skipped missing contrecord at' doesn't show up.
This machine is old, slow, and 32-bit bigendian. I first thought
the problem might be "didn't wait long enough", but it seems like
waiting for replay ought to be sufficient. What I'm now guessing
is that the test case is making unwarranted assumptions about how
much WAL will be generated, such that no page-crossing contrecord
actually appears.
Also, digging around, I see hornet showed the same problem:
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=hornet&dt=2021-09-29%2018%3A19%3A55
hornet is 64-bit bigendian ... so maybe this actually reduces to
an endianness question?
regards, tom lane