I think the main difference this time will be; that there will be no development time required for a separated drawing engine (with all of blenders modifiers and extensions, the pain was that everything had to made multiple times for each render engine. ).
Initially (old engine) rendering BI or Cycles would be way to slow to create a game, so there was BG with its own display rendering engine (openGL). But now that the render system gets an overhaul NPR rendering like eevee can be fast enough for games. So he doesn’t have to develop those parts any-more…
Also the physic engines allready exist (2.7x made use of the bullet engine, though not the latest version).
I don’t think baking (material?) is part of this, as Baking belongs to the cycles development team.
If you mean baking of physics then that’s recording of a Physics engine or recording interaction.
As for a game engine, to work it has to handle events, (ea object hitting or near object, or user input ), hence why i think it will be called interactive engine (as no longer it will be a separate 3d render system)
Node based workflows is verry popular these days, and there exists more census over standards then there was in the past (or at least limitation are better known). Learning from Blend4web and Godot and other game engines, i think such node system would be less difficult to make (the other projects already tackled the problems of working with nodes)… Though the goal seams higher, the new BGE could be used to export its logic to other engines. (sounds amazing, but not impossible, we already can make game props, so why not the logic)., Blend4web already uses a node system, but i guess it will be more of a ‘general common’ subset of such logic, so it could be exported to other engines (and refined there).
Hooking in other game engines might be an option as well, but if that was the goal, then they could have created just a plugin for unreal render instead of eevee, so i assume the devs set themselves for higher goals. Which doesn’t sound strange since 2.8 might the engine for the next 10 years or so.
It might also be a good time to start collaborating with the other opensource engines, i think they are allready “good friends”, i assume most opensource minded artist will use Blender plus something else, its a small world.
Also i like to note that Benoit has made a big surprise in the past with his BGE, if he is not upgrading forking the old BGE code then it means he’s onto something new something that will be a huge difference. And since coding skill don’t diminish over time, but tend to improve, i’m excited.
I’m trying to theorize and guess here so, @Joel_nl might be right as well.
Whatever the outcomes will be i’m sure we have exciting times ahead.