Weekly/Monthly record-keeping / maintenance?

[Jeff Blaine]

Every year or so, a user of ours reports a discrepancy
between on-disk usage for their spool compared to what
'FETCH' is reporting (as implemented via imapdu.pl)

Compact/Expunge via the client app buys nothing.

Clearly we are failing to do something as admins in order
to keep our Cyrus instance working up to snuff.

Can anyone shed some light on what we're doing wrong?

What are the regular maintenance tasks we should be
performing to keep everything working as precisely as
possible?

I found nothing useful in the Wiki regarding this topic.

# our imapd.conf with auth options censored
configdirectory:        /var/imap
defaultpartition:       default
partition-default:      /var/spool/imap
unix_group_enable: 0
sieveusehomedir:        true
autocreatequota:        200000
duplicate_db:           skiplist
annotation_db:  skiplist
mboxkey_db: skiplist
mboxlist_db: skiplist

# our cyrus.conf
START {
   recover       cmd="ctl_cyrusdb -r"
}

SERVICES {
   imap          cmd="imapd" listen="imap" prefork=5 proto=tcp4
   imaps         cmd="imapd -s" listen="imaps" prefork=1 proto=tcp4
   lmtpunix      cmd="lmtpd" listen="/var/imap/socket/lmtp" prefork=1
}

EVENTS {
   checkpoint    cmd="ctl_cyrusdb -c" period=10
   delprune      cmd="cyr_expire -E 3" at=0400
   tlsprune      cmd="tls_prune" at=0400
}