Great posts from Mike and Gavin in this thread. There's indeed no reason to panic over "too much centralization". Actually, setting an arbitrary limit (or an arbitrary formula to set the limit) is the very definition of "central planning", while letting it get spontaneously set is the very definition of "decentralized order".
Also, having fewer participants in a market because these participants are good enough to keep aspiring competitors at bay is not a bad thing. The problem arises when barriers of entry are artificial (legal, bureaucratic etc), not when they're part of the business itself. Barriers of entry as part of the business means that the current market's participants are so advanced that everybody else wanting to enter will have to get at least as good as the current participants for a start.
Removing the block cap means a hard fork, and once we decided to do that we may as well throw in some "no brainer" upgrades as well, like supporting ed25519 which is orders of magnitude faster than ECDSA+secp256k1. Then a single strong machine can go up to hundreds of thousands of transactions per second.
That's cool. Please core devs, consider studying what other hard fork changes would be interesting to put in, because we risk hitting the 1Mb limit quite soon.