riff_raff is right, there is a minimum fuel flow for an engine that dictates how quickly it can slow down in normal circumstances. This min fuel flow is what we see when we drop collective rapidly, or when we cut the throttle.
But when a rapid acceleration due to some failure is sensed, the engine has a means to cut its fuel flow very sharply, to contain the max rpm and prevent a burst. In the case where a large load is being shouldered by the engine and is suddenly cut, such as where an engine drive shaft fails, or a sudden internal transmission failure occurs and the engine is now unburdened but still producing gobs of power. The overspeed protection system (designed because the FAR/JAR requires that the engine handle a sudden cut like this) then senses the rapid upspeed of the engine, far faster than the normal fuel control. This sensitive system is virtually always electrical and triggers a cut signal that dumps fuel at a very high rate, forcing the engine to momentarily "fail" and stopping the upspeed before it becomes destructive to the engine.
The cut shaft is the worst case, since the engine sees virtually no load to absorb its power. Usually the overspeed sensing circuit has a way to judge the acceleration so that it is not just an absolute rpm protection, it drops the triggering rpm as the rpm rate increases. Most overspeed systems return the fuel to normal levels as the momentary event is recovered.
Here is an EASA document that describes the need for the test:
https://www.easa.europa.eu/system/fi...ilure_PUBL.pdf