Right, I get that it “cranks until it runs”, what I’m asking is how does the dme determine that it’s running? Is it looking at rpm’s meeting a particular threshold? Alternator producing a minimum voltage? Etc.

Of course, it’s really easy for a human to be trained to know when to let go of the key. As a Software Developer I am saying that it can be complicated to tell by sensor data alone since the situation can vary over time. I’ve had engines in the past that sometimes needed to be cranked for quite a few seconds because they were stumbling and not fully catching. That scenario would be difficult for a dme in my opinion. The dme has to weigh whether the engine can run on its own against running the starter motor for too long. I don’t think that’s an easy problem to solve so am trying to learn from others who might know, how the dme comes to the decision that it can disengage the starter.

Remember that the dme can’t “feel” the engine stumbling or hear how the exhaust sounds. It has microphones in the knock sensors but engine won’t be knocking at startup necessarily. It also can tell if a cylinder is misfiring but I don’t know how reliable that is at the low startup speed.