Recently we have been seeing a steady increase (nothing too crazy, just more than usual) in "Deleting corrupted cache record" log errors. What struck me as weird was that it would be for the same message over and over, indicating that the deleted record was making it back into the cache. Digging into one, I realized that the first error was actually this: Dec 4 19:08:59 b.imapd.sonic.net dovecot[2696]: imap(xxx): session=wnZqMH0oHOViI3NN Error: Mailbox INBOX: Maildir filename has wrong W value, renamed the file from /var/spool/mail/17/62/xxx/cur/1733363505.M428026P1491073.b.imapd.sonic.net,S=8966052,W=12155730:2, to /var/spool/mail/17/62/xxx/cur/1733363505.M428026P1491073.b.imapd.sonic.net,S=8966052:2, Dec 4 19:08:59 b.imapd.sonic.net dovecot[2696]: imap(xxx): session=wnZqMH0oHOViI3NN Error: Mailbox INBOX: Deleting corrupted cache record uid=6430670: UID 6430670: Broken virtual size in mailbox INBOX: read(/var/spool/mail/17/62/xxx/cur/1733363505.M428026P1491073.b.imapd.sonic.net,S=8966052,W=12155730:2,): FETCH BODY[]<11796480.393216> got too little data: 359249 vs 359250
Also worth noting that the emails themselves do have a highly unusual number of carriage returns. What causes this and is there a way to fix it besides deleting the offending message?