As noted, I switched hosting services a couple of weeks ago. The new hosts have 10x the memory and there are two of them for load balancing. Here's a graph (c/o Pingdom) showing "downtime" incidents:
Notice any changes circa October 10th?
FYI, the incidents usually lasted on the order of ~2 minutes and correspond to a server restart, usually due to the site's pixel crunching overwhelming the CPU limit on the servers under load.
Ω