Seen in the flatpak.
Kube immediately quits with a message of "Resource temporarily not available" which is the EAGAIN error code, which means the environment was locked by another process.
The problem eventually resolved itself as I tried to inspect it with sinksh and mdb_stat, not sure what fixed it exactly.
It's possible that the stale lockfile cleanup somehow depends on pid uniqueness which is not given running in flatpak, who knows.