So the idea is that if I'm an early-stage startup, I can set up a PFS cluster early on, store my logs there, and then I'll still be happy with that infrastructure 3 years later? Like the data pipelines that I write early on will still work, but so will the ones I add that join logs data with CRM records, experiment data, etc?

How hard is the extra step of setting up a PFS cluster? How much harder is it (how much code does it add) to append to a log in PFS than to a log in S3?