The market is tested, though. Auth0 is a notable player in this space and they seem to be making money.
Like you, I've been pretty unhappy with this class of product. Auth0 provides just enough to be dangerous; upload Javascript with no way to unit test it to gate auth (that broke!), support for multiple social providers but no built-in way to unify accounts (do that yourself), etc. They also have an insanely low limit for OAuth client applications; so low that we had to buy an enterprise contract to scale out to production.
(BTW, the way I tested auth hooks after an outage caused by a faulty one was to implement their API in Go, embed a Javascript interpreter in our Go tests, and then execute the hooks against an in-memory version of our API server. That eliminated any server-caused auth hook breakages. But I have to ask, why am I paying them when I have to do all the work?)
A competitor could be good here.