Is moderation a material part of Twitter's headcount? Like is that what 1000 people are doing?

I also have to imagine the graph and recommendation engine exponentially increases the infrastructure requirements and code complexity.

just running infrastructure at a certain scale is a really hard problem. You will deal with hosts going bad for arbitrary reasons (some hardware, some software), weird infrastructure problems (networking broken, spurious connectivity in routers, etc), deployments not being rolled out to all hosts in the same version, etc.

You can work around a few of those by having really good tooling, and by trying to offload to cloud providers which do a bunch of the work for you, but it still won't be easy. At the scale at which whatsapp and Twitter operate 50 engineers is nothing - in many products this will be less than the size of teams who just do some basic server rollout and operations.