My view on this, incase its not clear, is that cockpit video should (would) be treated the same way as CVR / FDR recordings. It's purpose is to supplement the CVR / FDR. So the data can ONLY be used as part of a safety investigation not in civil or criminal investigations. Now if the all concerned agree there could be a separate "recorder", actually removable media this could optionally be used for other purposes. This is the same as a Quick Access Recorder. The system fitted by Robinson does exactly this. It has a internal memory which can't be accessed by the user but will accept a plug-in memory which the user can use if they want.
For accident invetigation the camera is best sited looking forward to catch as much of the controls and instruments as possible. A bit of outside view good too. A view facing the pilot(s) is of very little value.
The big advantage over a CVR / FDR is that it tells you who or what caused the "input" recorded on the CVR / FDR. Examples from past incidents include:
A SLR camera between sidestick and seat.
A Sandwich in the hand
A NVG case as a control "lock"
A iPad in the controls
In these cases the cause was found, not always definitively, by other means. There must have been many accidents where the cause was completely missed in the past. And of course it will tell you who operated a switch..
I'm not sure how a cockpit camera can monitor fatigue unless it gets to the point wher the crew are actually having microsleeps. And in any case you don't need image recording for that, just some image processing software and a loud horn....