I've had this happen on a couple of other sites before as I was moving them from one hosting account to another, but today it really impacted one of my clients badly where the site had been in place on the new hosting system for two months. From what I can see, it seems the _site_transient_wpmudev_updates_data entry in the Options table is somehow causing some corruption that kicks off the mysql repair process and hangs the database so that all new requests hang until the server memory is exhausted and Apache crashes. This happened on a site that happened to be experiencing higher utilization (500+ page views today, which probably would have been closer to 1,000+ if not for the outages). The contents of the entry are over 300,000 bytes long. I don't know if it's a problem because of the size or what, but my solution was to do the following:
1) Disable the WPMU Dashboard Plugin
2) Delete the row
3) Successfully ran a repair
The problem was recurring every hour on the hour (more or less). When it occurred, the only solution I have (since I am using a managed VPS at A2Hosting) is to reboot the VPS... clearly if I had actual root privileges to the VPS, a recycle of the MySQL daemon would likely be adequate... regardless, it is highly impactful.
It looked like this row is used by the hidden Update plugin, but I have not investigated any further.
There may actually be something else I'm missing here, but the action list I provided above did solve the problem.
In case it matters, the table in question is using the MyISAM engine.
If needed, I have a dump of the row in question.
Ambrosia Web Technology, LLC