Upload & Sell: On
You'd also need to match the exact thickness of the original mirror.
Not necessarily. If the new mirror was slightly thicker or thinner, then the focusing screen would simply need to be re-calibrated. Since it is easier to add more shims, I would say that a slightly thinner mirror would be a better choice than a slightly thicker one. If the mirror is too thick, then removing all shim(s) might not be enough to make up the difference. The only recourse at that point would be to reduce the edges of the focus screen so that it is inset into the pentaprism slightly. I had to do that once with an E-330. Not pretty.
Either way, the focusing screen would have to be re-calibrated after such an operation, so the mirror thickness is not critical (but must be within reason, of course).
You would lose AF for sure, since the sub-mirror and AF sensor would be permanently in the dark, but you would not lose metering because it is in up behind the focusing screen.
Many have suggested that metering would be way off, due to much more light being reflected off the mirror, and would not be correctable with the 5D's meager +/-2 stops of EC. However, I am pretty sure it would be easy to deal with.
This is my logic:
1) When people replace the stock screen with an Ee-S screen, the 5D needs to have a custom function set to let the camera know that it has an Ee-S installed. This forces the camera to change its metering algorithm, which is necessary because the metering system is receiving less light through the Ee-S screen than with the stock screen (the viewfinder is "darker" for the same reason).
2) If a 100% silvered mirror is installed in that same 5D with an Ee-S and the custom function set to Ee-S, the meter will be receiving much more light than it should, resulting in under-exposure which may be beyond the 5D's +/-2 EC.
3) To compensate for the extra light reflecting from a 100% silvered mirror going through an Ee-S and hitting the meter, simply set the custom function back to Ee-A or whatever it is (not Ee-S).
In other words:
60% silvered mirror + Ee-S + C.Func(Ee-S) ≈ 100% silvered mirror + Ee-S + C.Func(Ee-A)
In this case I hypothesize that the ≈ could easily be compensated for by the 5D's +/-2 EC.