I'm beating a dead horse here, but did you ever try angling Golems as short term but ultimately unfeasible superweapons? For example, making them fairly beastly but saddled with a decently high global wave multiplier and a very long repair time so that there'd really only be a short window of opportunity for them to wreak havoc?
Not that specifically, but I can't see players really being overly enthused.
Not really, I've never been a big fan of "great power temporarily". The biggest reason I stopped playing Bioshock was because the developers actually let you take over the "big daddy" units (giant mech like robots that kicked ass), but only for about a minute, and only in rare cases. It was actually more heartbreaking to have such a powerful ally temporarily and then have him turn against you again, than it would have been to just never have the ability to "convert" him to your side at all. It's probably the reason I stopped playing that game.
I was also a big fan of Supreme Commander 1 and 2 (as well as Total Annihilation, the developer's original RTS game) and I can't imagine how much it would ruin the games if your "experimentals" were short-lived units that were only meant to be used once.
Some other places I've seen the kind of "short lived power units" are Dungeon Siege (not very successful as RPGs go) and World of Warcraft with the Warlock Infernal and Doom Guard, which were eventually both made into simple player spells after enough people complained.
In general, though it may be a good balancing technique, it tends to be a very frustrating mechanic for players.
I'm interested to see if a moderate energy cost (~150,000-200,000) and a small AIP Boost would give players an incentive to take them without feeling they were way too much of a risk. It being a new game mode and all, the pressure is a lot less on Chris to make the "perfect solution", and there's a lot more room for balance error since if people are unsatisfied with it, they can simply turn it off.