I can imagine Twitter has a lot of services that don't see a lot of traffic and a lot of people who don't have experience with high traffic services. Take your

I can imagine Twitter has a lot of services that don't see a lot of traffic and a lot of people who don't have experience with high traffic services. Take your calculator and do calculations along with me: 10ms = 1/100th of a second. 8 bits = 256 possible values. For a given node you have at most 256 possible IDs every 1/100th of a second or 25600 every second. This is absolute limitation which means, you can't just straight use the library without some workarounds to get more traffic through a node. You could of course do things like assign more hardware addresses to each node (maybe even messages on one and odd on another hw address) but then you soon find out the workaround is more complex than just writing it from scratch.