The DB write spikes are very likely to be EBS glitches, we've seen them do that frequently.

The crazy CPU issue might be related to the EC2 bugs in the Lucid Lynx kernel...Maverick is also problematic but Natty Narwhal has fixes for many of the weird Ubuntu-on-EC2 issues

Nice - interesting that EC2 instances vary so wildly in their performance. Will Amazon ever attempt to spot and 'repair' sluggish instances?

They haven't so far. Even some of them are designed that way.

Designed to be sluggish?