There is at least one good reason, one which at the least I and Peter Todd support. It's an authoritarian one, though. (And thus, as above, I'm not really for a limit.)

Necessity is mother of all invention, and there are a shit-ton of things than can and are being made to aggregate transactions in different ways, ways which have different incentive, economic, and security implications. This impending limit has caused a great deal of thinking and research on sidechains, treechains, and several other things.

For this reason, while I'm for removal of the limit, I'm in no great hurry to see it happen, until TT1C (time-to-1st-confirmation) begins to be detrimentally impacted. So I've stumped a bit on the "leave it alone" side in some places.

So why am I here? (1) we're nearing where it could start non-negligibly impacting TT1C soon; and most of all (2) so long as this is getting debated hotly, I must insist that if it is to be changed, don't just replace 1MB with another dumb limit!

edit: So in any case there's no good reason for 1MB forever, which was your phrasing. In fact there's no good reason to have a blocksize limit while having a mintxfee; they solve the same problem.