On a related note of using a existing prototype chassis as an F1 engine mule. In the early 1980's Porsche installed the Porsche made TAG V6 turbo into a 956 prototype.
Plenty of speculation as to what Honda's issue is, but some say potentially ICE vibrations that cause problems with the MGU-K (which is directly driven by the crankshaft per regulations). All of which seems to be driving some type of electrical insulation failure. You would ask why this wasn't discovered in internal testing? Maybe it is something related to the dynamics of it being installed in the F1 chassis?
Then if you wonder if that exact issue may have shaken out in mule testing, you should ask questions about how close a mule can be to the F1 car. It would have to have the McLaren transmission (I assume McLaren makes that and not Honda), the correct suspension loading on the engine as a stressed member, etc. Then you wonder if the weight of a prototype might not match up well. I don't now how much ballast is placed in a typical LMP2, but the class runs at ~930kg while F1 is at ~700kg (25% lighter). Lots of variables that in some ways they can control better on chassis rigs (other than g forces). I also expect dynamic simulations in dynos (including full simulations of each circuit they race it) should the norm for helping to get baseline engine maps vs. having a test driver pound around a single test track.
Here is the oft shown example that Porsche uses...
https://www.youtube.com/watch?v=fv53RbvgfGc
It can only however simulate 1G which would seem like enough for oil pickup, but it doesn't simulate things like de-foaming ability of oil in a tank at high g levels.
Richard