I'm not going to lie, I had a good laugh thanks to this post. Not that I fault the method being used to combat memory leaks. Depending on how much (of pg's) time a restart takes, this may be the most efficient way of handling it, but the idea that the site where we all discuss the finer points of multi-region-zone-datacenter replication redundancy just gets restarted when something isn't working correctly humors me.