If anyone here believes this, make your best estimate as to how many man-months you need for a re-architecture and how many hundreds of dollars you'll save in hosting costs, then do the division to get your effective hourly rate. I'll pay you that plus 50% for contract programming work.
For $21.6k/yr I'd say it's worth a week or two re-architecting.
Yes, it's a different game if you're profitable and $21.6k is negligible, but if you're a startup you should be spending time to optimize things.
The other point is one of scaling. If you're paying $2k/mo to support 15k users, when you scale to 15m users, you could be paying $2m/mo unless you fix things early on.
It looks an order of magnitude too expensive to us, there's probably some simple thing wrong in the architecture.
Maybe we're all missing a key complexity of the service. The Archival service might be it.