From my experience you can't even tell from careful study of most recent datasheets about all properties of a particular chip - there's always things like silicon errata and esp bugs in datasheets - quite frequently.

So if after careful comparing 3430 with 3630 it looks like they might be drop in compatible (not really unlikely a result), you got exactly one chance: try it. You "just¡" need a (few) OMAP3630 SoC chips incl PoP storage slice of your choice, and two or three N900 for that, and a really gifted reworker as I found them in Taipei, where they actually achieve yields of maybe 20..30% on swapping that chip formfactor, by using tools that are in the $10k range, maybe even X-ray for 100s of k$


The question is: who would want to spend such MUCHO DINERO, to pimp a rather obsolete hardware like the N900 (yeah it is, in fact!) by such a little as 1GHz instead of 600MHz. Even the other improvements don't really bring your N900 just in vicinity of a HW you could buy for the mere cost of sourcing the 3630 alone. Get an iPhone7 and be happy! You probably can bribe someone at apple R&D for the amount that this enterprise of a pimped N900 costs, so you might actually get a prototype iPhone7 right here and now ;-P

As mentioned before, that's such an utterly useless and unproductive, moot discussion.
I suggest next time we argue and call names about whether or not you can craft a V8 engine from a solid block of 9999 gold, or if that's doomed effort. Sounds quite as reasonable as this debate here.

abill_uk is right, it might be possible (duh, I really had to say this). The ones with concerns are right, all the mentioned issues might place invincible obstacles in your way, and chances it will pan out are minimal. I know I'm right about the uselessness of the whole debate (except if you're willing to spend $$.$$$ to prove you're the mad über-geek - a questionable honour)

/j