> even at the extreme where it reduced time to market to 5 seconds.
If this is your impression of internal tooling, then your internal tooling team is terrible.
The places I've worked with good internal tooling teams provided a dramatic decrease in time to delivery. Automated failovers? Don't have to write it, it's in internal tooling. Automatically rebalancing traffic across datacenters? Already in the framework. Service discovery? You already know.
I could go down the list, but with a good internal tooling team, the only thing I've ever had to worry about was my app's business purpose. I don't have to muck with configuring monitoring, or load balancers, or getting authentication to talk to another app, or how to get a database, or etc.
The difference in time to delivery is dramatic.
Places I've worked without internal tooling take like 2 weeks just to get infra for a static HTML site.
The places I've worked with bad internal tooling take 3 weeks to do the same thing, because you'll spend a week arguing about whether the tooling works.
The places I've worked with good internal tooling take like 15 minutes to do it. I spend longer reviewing an MR than I do deploying infra.
Tooling is a huge part of the reason companies throw fistfuls of money at clouds. They could get compute on-demand cheaper from Leaseweb or OVH or hundreds of other companies, but they don't. They pay the premium for AWS or Azure or GCP, and it's because of the tooling.