Interesting that you pointed that out. I was thinking the same thing when
I posed the question but thought that two tumbling AIs would pretty much
guarantee that the pilot would not recover.

I agree with you - with two tumbled AI's and no TC, recovery is
impossible. With a working TC, it is merely highly unlikely.
Therefore, if the panel space is there, it might make sense to keep the
TC. Not much sense, mind you, because if you manage to screw up badly
enough to put the plane into an attitude that would cause both AI's to
tumble, well, I'm willing to give very good odds that you're not going
to recover on the TC.

However, let's say having both the TC and the second (electric) AI is
not practical (probably due to space considerations). Would I rather
have an electric AI, or the TC? I would still prefer the electric AI.
First off, with dual AI's next to each other, I believe that following
a dying AI into an unusual attitude becomes far less likely, and thus
while the chances of recovery from the unusual attitude are reduced
slightly, the chances of encountering it in the first place are reduced
dramatically. Not so with a backup vacuum - you have to engage it.
And even if you do, half the time (in my experience more) the problem
is the AI, not the power source, so backup power for the AI does you no
good.

The problem with this analysis is the reliability (or lack of same) for
electric AI's. I've heard the affordable ones are not good, and the
good ones are not affordable.

Finally, there is the issue of training. If you have dual AI's with
independent power sources, it makes sense to skip partial panel
training. If you have only a single AI, even with redundant power
sources, that's not the case. In that case, a standby vacuum system
seems to be an unjustified expense - the money spent on it is probably
better spent on recurrent training.

Michael