IC: Super D&D
I kind of disagree on the point of progression. I could wax philosophical on the point, but in short, I think that mechanical progression is often taken as collateral in lieu of narrative or internal character development.
Another point is that a PbP game is pretty slow burning. If you're counting on XP to carry your PCs up a tree of feats, you may not see your characters develop "super" powers for several RL years. Most Supers games start with your characters effectively already advanced, with powers fully fledged, specifically so you don't have to wait through three levels of "Ray of Frost" before you get to play Iceman.
The main bugaboo that I would have with a system similar to d20 Modern's Talents is opportunity cost. If the talent trees really do develop into genuine super powers that have enormous mechanical effects, there will be an enormously high opportunity cost to take colorful, diversifying, but less powerful effects than, say, FISS+B (Flight/Invulnerability/Strength/Speed/+Blast). Since Talent resources would only be available as characters "level up," they are effectively required to meet performance standards to face appropriate challenges. This means that if they didn't pick talents that enhance their abilities in a mechanically relevant ways to meet those challenges (say, if they wanted to branch into a new type of talent, or they are maintaining a varied power set over multiple branches and can't "go deep."), they become increasingly mechanically irrelevant.
This is very similar to making "bad" advancement decisions in 3.5, and is an issue inherent to level-based advancement. In a level-based system (rather than a free-points system like M&M), you have to chase a mathematical curve with very limited resources, and this can result in some very unsatisfying choices that don't really make your character any more interesting. (See: Spell Penetration).
I personally feel that that problem could really be exacerbated by making access to cool super powers like at-will flight and cutting mountains in half dependent on staying on the bleeding edge of talent trees. Or worse, characters could gain those incredible talents and then just have to let them mechanically atrophy into irrelevance just from lack of resources to commit, owing to the rapidly scaling mathematics of 3.5. (See: Turning and Cleric PrC's)
Some people do consider that a feature, though, and not a bug. Rewarding specialization and enforcing niche protection is something that point-buy systems (and especially games like M&M) struggle with, without cooperative and communicative players and GMs who really understand the systems and their pitfalls and are willing to write weaknesses into their stats.
Concerns of that nature would push me towards True20/M&M over trying to hack D&D, if powers are a central part of the game.
However, if it's leaning more towards D&D with everybody having a few spell-like abilities or extra bonuses, that's a much simpler add-on.