That's actually normal now. We've come full circle, but it's called "observability" - since you can't integration test the 100 microservices that you have (for no reason), just throw it into prod and see of anything turns red.
I'd suspect it should be possible to test those 100 microservices. If you can do something, then you can do it twice, especially if your automation is good. If you can't, then maybe some of the systems came together in a non-reproducible or undocumented way and it's time to clean up a little. If you know your traffic you can build load tests that simulate production use.
You could find a bug in pre-prod and live edit it in prod