ran into a dumb limitation with google drive for desktop and cloud code. i had an apps script pulling my email threads and dropping them into a google sheet on a schedule. the theory was that since drive for desktop syncs everything locally, my cloud code instance could just read that sheet. except google sheets don't actually sync as real files. drive desktop creates these tiny pointer files for anything in a google native format, so a gsheet file on disk is basically useless to anything except the drive client itself. cloud code runs sandboxed and can't resolve those pointers.
the fix is simpler than i expected. instead of writing to a google sheet, have apps script create an actual csv file in drive using driveapp.createfile. regular files like csv, json, txt sync as real content to your local disk. so now the pipeline is apps script dumps email data to a csv, drive syncs it down, and cloud code reads it like any other file on disk. if you need the google sheet for viewing too, you can have a second trigger that exports the sheet contents to the same csv file. only caveat is if you're on drive's stream mode instead of mirror, the file might be a zero-byte placeholder until first access, but if your setup already resolves drive paths then you're good. this one change basically solved the entire problem.