We have a customer that had their Exchange store crash. Normally we can restart or remount the store fairly easily, or at worst restore from backup. It didn't work for this customer, so I had to do something more drastic.
I used the eseutil /p command followed by the store path to repair the databases. I don't normally use this because it could damage the store further. Exchange 2003 is just a big Access database, unlike Exchange 2007 which is SQL.
Access is not very robust when you start messing with links from email to headers, but in this case it was the only solution.
After running the repair I had to delete the E00 files and then run eseutil again with the /d switch to defragment the database. It really fragments a lot when you run the /p, so this is to speed the store back up again.
Everything mounted great after that, but the next day it crashed again. I looked at all the log files and noticed just before it crashed that the backup program was running, so I disabled it and fixed the store the same way again. It was all good for five more minutes until it crashed again. Then I suspected a hard drive sector problem.
This time I defragged the hard drives, and then rebooted and ran a scandisk. It found a lot of errors. After it rebooted I ran the fix again, and this time it stuck. No more crashing.
But the hard drives are on their way out so we had to sell the customer a new server because this one was many years old. Planned obsolescence is an awesome business model.