What was the specific metric that forced the move? (Bandwidth, Function execution time, Cost?)
How much 'DevOps' time do you actually spend maintaining the VPS vs. building product?
I want to avoid premature optimization, but also don't want to be locked into a platform that will bankrupt the project if it takes off.
This way you get away from the vendor lock-in. You should be able to migrate providers in the future without too much work and have a clear path towards hosting on a VPS or your own servers when you get enough traffic to warrant it.
Effective cost should probably fall somewhere in the middle, too.
But you won't have to deal with most of the ops headache that comes with the VPS approach: OS updates, firewalls, filesystems, startup orchestration, log fike rotation, etc.