sonarr keeps track of which torrent hashes its already downloaded. unfortunately it appears to get checked at the import stage - and when that happens sonarr just skips the import and deletes the incoming files completely wasting all that time and effort.
i would like to suggest that the check is front loaded, dont use the sonarr label use sonarr-manual or some other (or no label) on the download job so that sonarr doesnt even try importing (and then pointlessly deleting) the files when the download has completed.
that way we can get to it later ourselves and manually import them.
yes, i know, why doesnt the user just go and do it themselves later? counter - why exactly should they when its perfectly reasonable that sonarr just does it itself? that is the point of automation is it not?
plus, if ive lost the shows from data loss (which i have) and im trying to redownload them again - sonarr doesnt give a crap - it still deletes them even when the files are missing.
its dumb, and it could easily be better.