I have had a quick look through the several new issues arising from the #AIslop that was put into #rsync 3.4.3 by its creator, and one commonality seems to be that the 'vibe code' did not bother to account for building on systems that don't exactly match the current Linux openat() and mknodat() system calls.
https://aidirtylist.info/citations/andrew-tridge-pushes-flurry-of-code-slop-to-rsync-respository/
This has already caused a whole bunch of extra work for real humans. e.g.:
https://github.com/RsyncProject/rsync/pull/925
The major change is exemplified by this, but there has been wider work across several syscalls:
https://github.com/RsyncProject/rsync/pull/887
That itself was spurred by this AI slop from rsync's Debian maintainer, who took an earlier commit as a green light: