...one other possibility is an AFCS malfunction where the rad height didn't capture at the end of a TD. (or possibly finger trouble)
From the ADSB it looks like altitude (rad alt) was captured, some power pulled in which sent the airspeed back to 100, then power back to slow down. What AP modes are coupled here? Who, how is the power controlled?
Reminds me of an Aberdeen 332L flying a scuddy non-precision to a beach airport that ended up in the water 10 years ago. Again curious on the training and SOP for the 92 in this situation, and if Bristow, CHC or even Cougar match up. We've spent over 20 years trying to figure out the best integration of autopilots in helicopters, it should be figured out now.