Since I had my big issue with a bad controller and an inconsistent ZFS Array, everything is running well again. I did several scrubs. But due to the inconsistency some files seem to be damaged and irrecoverable. To have clear stats, is it possible to grep all file paths and just delete them, to have a “clean system”? Resetting the stats wont help, since in the next scrub the list is there again. What do the ZFS Wizards recommend?
TiA.
Don’t delete files, maybe only one damaged block in file.
zpool clear <poolname> will clear the errors. Not the files.
With piecestore it was pretty common for a handful of small .sj1 files to show as damaged: just delete them. With hashstore if it’s the larger data log files then rebuild them.
How to do this? 
I don’t think you can. If the hashstore is damaged you can rebuild it from log files. If log file is damaged itself — then just leave it alone, you will have data before the corruption intact. And there is some chance that after corruption too (depends on log file layout which I did not bother checking)
Either way I would not delete anything.
You are proposing losing more data to have an appearance of “clean” system..
There is a Go Script for fixing corrupted Log-Files
write-hashtbl dont fix log files it fix only metadata.