Just fixed the issue on my system.
Looks like when writing to a asynchronous file system (async NFS) Picard will get confused. It’s writing the file and then immediately trying to reading it to confirm it was saved. It’s doing this too fast and on slower storage systems the file has been not be committed to the disk before Picard tries to read it. So it gives you a file not found error.
This explains why I was able to fix it by forcing synchronized transfers on NFS. I am now making Picard wait (in I/O wait) for the write to complete before the read can be executed.
So this may be an issue with network storage protocols. An easy fix might be to add in an option to delay the confirming read by x amount.
Or I’m totally wrong.