I only caught the match in text, but if Gambhir was dissatisfied with his home pitch, he may have had quite a surprise in Chennai. Delhi's planning has been a bit questionable- they have erred in persisting with Warner at the top so long, even with both Sehwag and Gambhir back together. Consequently, all three are struggling for runs, and that's weakened the once-formidable top order. A revert to the regulars Dilshan and AB in the middle would help them. There's also an alternative of beefing up the bowling by playing Nannes, McDonald and Vettori together- and Collingwood is an under-rated T20 seam-up bowler.
Mithun Manhas is a bit of a surprise package for this team. He's done well as a steady partnership-builder in low-scoring matches, but lacks striking power for the high-scorers. He's built a valuable partnership for his team, and may be best suited as the man at the other end of the crease, rather than the more glamorous striking end. His less-experienced counterpart Kedar Jadhav, one of Maharashtra's new breed of T20-ready cricketers, may be more useful at either end, but needs to be more consistent.
Then there's the question of choosing between Umesh, Sangwan and Yo Mahesh. For some reason, after one bad match, Umesh (fair, he's not quite ready for T20) was benched for the very average Sangwan, who's stayed there after one good game. He's taken his chances, but the quality isn't really there. They can continue to invest in Yadav or revert to the more useful Rajat Bhatia for their full XI- there's no way Sangwan or Yo Mahesh can add much to their places.
As for Chennai, they've got it wrong on a few counts. Bollinger's hardly the ideal T20 bowler, and his pre-IPL stats suggest so. I wouldn't be too critical of Dhoni coming in ahead of Raina- he's a better pick at three, being the smarter batsman, and he's done little, when he sends Albie Morkel ahead of himself. The bowling changes may be a bit questionable, as he's not getting much out of Balaji, who is one of India's better (albeit expensive) T20 bowlers, while the lower order, though productive, hasn't been tested.