I'll add in my 2 cents, for what its worth, there would be quite a number of problems shifting to OpenGL at this stage anyways. For a start, the dimensions of most bitmaps right now are weird sizes.
So even if you were to change all that and make all bitmaps powers of two, it will be a pain in the ass for everyone else.
If you were to load bitmaps, then cut them up in memory to power of two chunks, you would wind up with a lot of texture handles (Some of the large tilesets may produce a hundred or so texture handles!) and this isn't really an efficient way of doing things.
Texture sampling is also going to be an issue. Ideally with the way OpenGL works and so forth, you will probably see a lot of bleeding issues. I generally give a 1 pixel border on a lot of my OpenGL resources so that I don't get this problem, although you can some what fix this by using different sampling methods.
I don't believe the switch to OpenGL is going to add much, and will probably be a big gigantic head ache for you. I can understand why, you may have liked to do it ... but certainly it would have been better to have done it from the start.
My next 2D engine for my game (which is still in documentation planning) will be using OpenGL however, so feel free to chat to me about it at any time!