Hello. So, like eyeliner’s pointed out, the integration with the BGE is really a nice feature. The fact that you don’t need to export and import models, as well as being able to quickly play the BGE to see the game unfold are great features. While you might have an idea about merging in other engines to Blender, I don’t see a need to get rid of the BGE. If that is really the preferred option, then some Blender developers should work on integrating other engines into Blender. After all, the BGE is a part of Blender - what concerns it should concern the Blender developers. Otherwise, why leave it to the few dedicated BGE devs to work on integrating in an entirely unfamiliar engine (unless they wanted to)?
EDIT > As a side-note, the BGE’s not that far behind. It’s not exceptionally fast, but with just a few improvements, it might literally be the easiest and most powerful open-source indie game engine. There’s nothing wrong with Blender, and there’s nothing wrong with the BGE - it’s just not getting enough developer attention. If you look at, for example, devmaster.net, the BGE has a somewhat comparable score to Unity (4 stars with 33 reviews vs 4.3 stars with 67 reviews). < EDIT OVER
There’s already been an attempt at making a separate, open-source game engine that’s integrated into Blender, which is GameKit. The problem is, it’s not nearly as easy to use as the BGE, and it’s not as advanced in terms of capabilities yet (as far as I can tell). Also, the integration itself isn’t complete (I think stock Blender GLSL shaders aren’t imported, for example). Integrating an engine that was by no means made to import data from Blender would probably be more difficult.
Anyway, there’s been other threads like this before that had pretty large portions of the engine’s regular users comment on it. You can search for them / go back a few pages and find one.