> Docker uses overlay, the underlying filesystem is ext4 Does the DB file, or its paired -journal file exist on the host filesystem before docker spins up? And are both threads running in the same docker instance? (I think they belong to the same process but just checking) > I can only reproduce this by running the same test a few times (sometimes the first one is enough) on a rather powerful machine that's usually used for our CI. Not sure if that has to do with its speed or if it's related to CPU features. For what it's worth, I ran all tests in a docker image (and can't do otherwise on that machine) Can you reproduce the issue if you run the same docker image on a different machine? Does the app perform further DB operations (in any thread(s)) after hitting the error? Either explicitly or as part of the runtime's cleanup. Wrapping your app with `strace -f -e trace=file,read,write,fsync,fcntl,fcntl64` might be informative.