Detecting engine failure isn't trivial - a failed engine can still spin pretty fast with airflow going through it.
Modern predictive techniques could easily detect engine failure (ie. "Here is the history of the last 10 seconds of the throttle position, and here is the current RPM, is the engine failed?"). But aircraft designs don't like to use such techniques - they prefer simple thresholds. And with simple thresholds, a bunch of engine failures can get missed, especially partial failures.
I think the is a big issue across aviation. The system just gives up and dumps the problem on the pilot who then has to parse conflicting data quickly. I wonder if it would be better to treat this more like a navigation problem. Fuse together different data with known error levels and show the pilot the result. And include more solid state sensors like GPS, gravity sensors, and cameras in the mix.
Modern predictive techniques could easily detect engine failure (ie. "Here is the history of the last 10 seconds of the throttle position, and here is the current RPM, is the engine failed?"). But aircraft designs don't like to use such techniques - they prefer simple thresholds. And with simple thresholds, a bunch of engine failures can get missed, especially partial failures.