I brought timescale (a postgres extension that brings time series to postgres) to my current IoT job, and it has been the biggest engineering win of my career. The tooling, community, extensions, and quality of postgres + very high performance inserts + reads has reduced costs for the business dramatically versus the previous cassandra + sql server setup. We have tens of billions of events in multiple timescale hypertables, with many thousands of inserts per second running on r5.xl nodes in k8s (4 vcpu, 30gb ram) and they don't break a sweat.

So yeah, it's "good enough".

Could you send me an email because my company is starting something similar

I realize now in after hand that we're talking about two different things. You're mentioning time series data, I talked about bitemporal databases.

Right, and it's probably "good enough" for most use cases. But, as always, it can't cover _everything_. I'm just making an argument against the sweeping argument that "postgresql is for everyone and everything"