This sounds like "yeah, it has bugs that need to be fixed", not fundamental reasons not to merge it. (Though I think they should've tested with KASAN+KUBSAN before submitting.)

Sure, bugs can be fixed. The bugs are one reason not to merge this iteration of the patchset. I don't dispute that it could eventually be polished into something worth merging.

(The second bug in the email linked above is a NULL ptr deref during an in-order recursive directory copy into the volume. That is real basic stuff; I don't think it required ASAN/UBSAN to hit. It shows a lack of even relatively basic testing, if I understand correctly.)

If a tester can't simply create a volume, mount it, and copy a directory tree into it without encountering a null pointer dereference, this to me raises serious questions about how this driver is tested and the lack of care in building it.