OK...
1. Change of data source causes mode deactivation.
2. loss of mode causes unintended flight path
It seems the crew were distracted by the abnormal, so the flight path was not monitored, and therefore deviation not spotted until GPWS. Obviously I think the CRM fell apart in regards to this.
The report also refers to them not noticing the mode deactivation or monitoring the FMAs. Now, is it being suggested that the mode deactivation should have been spotted as a matter of routine scan? if so, I disagree. Yes, in general routine scanning should pick up mode changes, I don't dispute that, but in this case the mode change shouldn't have come as a surprise.
I'm not familiar with the Q400, either it's avionics or checklists, so the following terminology may be wrong, but the gist of what Im saying should be valid. Comparable aircraft I have flown have had similar behaviours. I.e. Changing data sources might result in a mode change. Thus, changing a data source is an event of equal gravity to making a flight mode selection.
So it should have a thought process to match:
What modes have I got?
What should happen when I make these selections?
(make the selections)
Are the modes as I expected? (if not, why?)
Is the other pilots mental model of the current modes the same as mine? If not, which (if either!) of us is right?
In this case the very act of reaching for the air data selector should have triggered the knowledge that it would disarm the alt capture, and a wise PNF would verbalise that knowledge IMHO. 'this will kill your vertical modes' or similar.
So to me this doesn't just speak of a crm breakdown, but also weak initial tech training and/or poor recurrent training. Surely an ADC failure should be part of the recurrent sim training?
Please note I'm pointing the finger less at the crew than the training, but also at the industry in general.