[mmt]

> What factors could cause this to be a bad idea? Need for proximity? Network speeds? Security?

In short, yes to all.

Specifically, fallacy [1] numbers 1, 2, 3, 4, and 7. Maybe number 5.

None of those are necessarily insurmountable. However, given how relatively lightweight a connection pooler is to operate, especially compared with Postgres itself, it doesn't seem like an attractive target to "outsource".

[1] https://en.wikipedia.org/wiki/Fallacies_of_distributed_compu...