I'm in a bit of a pickle.  MySQL errors come from lack of memory.  This can be solved by limiting the number of threads Apache fudges out to handle requests.  However, by doing that, the server can handle fewer requests at the same time, meaning that extra requests will be placed in a queue.

What's odd with the downtime is that I'm assuming there's far more requests being processed than usual.  I've been monitoring things lately, and have noticed some odd things, like a single IP with 12 concurrent keep-alive connections.  It looks genuine (looking from the requests) - it's probably produced by some odd optimizing tool or something.

mod_evasive has blacklisted 3 IP addresses for possible DoS attacks, however, it doesn't seem to explain the downtime... >_>