1. I'm still here, and still writing p2p code pretty frequently, oddly enough.  Expensify's core database uses a clustering replication technology I wrote, and I make steady updates to it.  Basically, it uses a paxos-like distributed consensus algorithm to elect a master, synchronize all commits, then a two-phase commit transaction system atop that.  It's extremely stable and scalable, using sqlite as the storage engine (which is amazing, BTW).  I'm planning on own sourcing it, but haven't gotten around to it.