CJ - I really like that idea in principle.
However, I wonder if the accounting for it would be too onerous. I mean, you have a summoning with Stats X. Every round, those stats change, plus you get the obvious change of actually taking damage as well, plus any status effects that get tossed onto the summoning.
It's just another step in something that tends to suck up a fair bit of processing power at the table. Never mind the poor DM who might have to track three of the darn things during an encounter as well.
Cool idea, but, I'm not sure if it's practical.
Well, boiled down to a statement of principle, the idea is: Balance is most important on round one. It gradually diminishes in importance as each round passes. When the fight hits the stage of mop up or boredom, balance can rapidly fly out of the picture, with very little trouble to anyone.
Obviously, then, strong effects would start weak and get stronger over time. This need not be limited to wizards, but does have handling time concerns such that it needs to be limited across the fight as a whole. The first thing is that if the effects are designed correctly, there are diminishing returns for unleashing too many of them (except in the most important fights, where the extra complication might be acceptable anyway). That is, if the wizard already has his "Nuclear Effect" building up, then someone else chipping in might be overkill. Plus, the more PCs that are setting such effects into motion, the less PCs that are hitting hard
now. That's a key balancing concern, as pemerton noted.
One of the big ways I would keep the handling time under control is have the ongoing strong effects require some portion of the action economy, but not an "every round must concentrate" level of attention. Rather, build in interesting choices that make unleashing multiple such spells from one caster not generally worth it.
For example, with the animate blade effect, require the caster to use a minor action to sustain the spell, and a standard action to redirect all animated blades to new targets (not necessarily the same ones). A blade is getting animated every round, the spell is becoming more and more powerful, but while it is first animating, the caster will probably prefer to get a few other spells in until it builds to a critical mass. Maybe he starts it on round 1, does something else on round 2, and then on round 3 sics the current animated blades on separate targets so that they are doing a bit. By round 5 or 6, the wizard might prefer to spend all his time getting all the blades on one or two targets to take them down.
That's just an example, and would depend heavily on the details of the action economy. Another way you can think of how it should work is to compare to other theoretical systems. One way you could balance uber effects is to make them take multiple actions to cast. This is generally avoided in D&D, because it is boring--the wizard player sits there, "I'm still casting (sigh)," round after round, hoping to launch the game winner. The escalating effect does much the same thing, but gives the player something to do while it builds. But like the long casting time option, it should not fully free the character up while building.
You
could have delayed effects spells, where they are launched early, allowed to build on their own, and then require another action to grab and direct once they are built sufficiently. That could be another interesting decision. A souped up fireball builds 2d6 per round, to the maximum of the spell (higher than normal), and can be launched at any time for another standard. If a great time to launch it happens 1 round before max effect, the wizard might decide to go for it. Or if it maxes out and no good targets, hang onto it for another round or two.
