This looks like it might be the snapshot_of corruption XCP-ng docs describe: https://docs.xcp-ng.org/troubleshooting/xapi/snapshot-of. Interrupted operations can apparently leave that field pointing at the wrong thing, which is what makes a real disk get read as a snapshot. The docs say a fix is being worked on across several storage backends but don't mention XOSTOR by name, so I can't promise this is the same bug rather than something specific to the LINSTOR side.
There's a snapshot-fixer.py script mentioned there with a dry-run mode, it shouldn't change anything, just tell you if it sees the same corruption. Worth trying that first before anything else.
I'm not deep enough in XAPI or LINSTOR internals to know if XOSTOR is covered by that fix or not. If dry-run comes back clean and this really does look XOSTOR-specific, might be worth a mention to @Team-Storage, since production VMs are affected here.
