Recovering from crashes - improvements needed!

[Iain Stevenson]

I had a system crash which, as it turns out, trashed mailboxes.db and 
deliver.db.  Whilst trying to sort this out, I found that ctl_cyrusdb went 
into a loop.  I'm sure it's a loop because 16 hours seems a bit too long to 
repair a system with 5 users and strace showed no change for the 
ctl_cyrusdb process.  It would be nicer for ctl_cyrusdb to fail with a 
suitable message if it found mailboxes.db was garbage.

I also took the opportunity to change the mailboxes.db to skiplist. 
However, skiplist moaned about not finding a "skipstamp" - what is that, 
and how should I create it (if I need to)?

  Iain