Running a #ChiFS Share with more than a million files on a Pi isn't the greatest success. It works, CPU and memory wise it's perfectly fine, but writing out metadata for so many files to a shitty flash drive is very slow. Might need to consider a database format that is more efficient on I/O.
Actually, I just discovered a race condition in my code which could cause those errors. Perhaps I shouldn't be too quick to blame the hardware, after all.
We are a cute and loving international community Ｏ(≧▽≦)Ｏ !