SpaceX management is currently oriented to both high quality and high production rate. When their commitment to quality slips, such as traded off for production rate, quality will slip. When quality in a rocket engine slips, they explode.

When that starts, we may expect they will start by exploding mainly on test stands, instead of vehicles.

How many launch failures have been attributed to materials failures in the engines themselves, and not improper process and/or maintenance before actual launches?

My intuition is that the latter would exceed the former, and that test stands aren’t a realistic environment for predicting their likelihood.

Probably the main difference from operating bolted to a test stand is vibration both from its own operation and from all the other engines operating at the same time, and instabilities in fuel flow caused by the same. The bottom end of a launching rocket is a really hostile environment. Just the noise would kill you. Top is marginally better, depending on what happens at the bottom.

It is hard to imagine how you would determine, after the fact, whether a rocket blew up because of material failure or something else, but they seem to do it, anyway during development when they have a zillion sensors attached logging everything in real time: "hmm, a millisecond before the explosion, this reading went out of tolerance, and then this one, then this one, and then the data ends."

There are Falcon 9 launches carrying humans on crewed Dragon launches. SpaceX carries people to orbit and to the ISS.