Honestly the biggest reason we had to re-platform to the JVM was latency (fail whale was a 5s timeout) and not growth. Ruby was just too slow for all the stuff we wanted it to do. It ended up being 10x fewer servers AND 10x lower latency.
I feel like this is a pretty important comment that busts a popular myth about Twitter and Ruby. The notion that Ruby was a poor choice even for a young Twitter that hadn't hit scale makes Ruby (or at least, late 00s/early 10s Ruby) look even worse than it already does.
