I vote maya. Not because i use maya (I'm completely agnostic when it comes to software) but because it would open a huge market/user base for Corona meaning better business which will translate in a better renderer. Maya is still the most widely used tool in film and tv broadcast and that is not changing just yet. Autodesk seem to be focusing energy into Maya first, max barely and xsi .. Xsi has a dark future ahead. Modo by itself was not doing too well but the fact that now joined forces with nuke will probably make it a bigger name in the film industry. A version of corona for modo ( hey, even for nuke! ) could be very interesting as well. It just sounds better business to focus on these platforms first..just my opinion.
Octane is an interesting case as a renderer ( whoever said it is slow should consider a decent gpu ). It has a plugin for most 3d apps but it's functionality is really poor. It might be a good model to be everywhere and develop the tool later but it feels like neglecting a bit the core of it.
Arnold is completely the opposite. Their focus, as stated on their site, is to give the best possible support to few clients ( hence the licensing exclusivity ) first in order to grow the tool in a solid manner. And they only support xsi and maya.. These guys are aiming to film industry use mainly
Btw.. Whoever said arnold is slow for interiors should really try it more before saying that. There is more to rendering than just speed. ( arnold is damn fast overall )
I do think Corona could easily fit a bigger role than in architectural rendering or advertisement ( with all due respect to those fields ) . In the little time i've used it i found it scary close to what Arnold offers and that is a lot of ground for an alpha. If things like pmedia and exr2 ( deep data multi view ) get implemented Corona could become a really solid player in film/tv
Heck! I would even invest on it! :)
My two cents
.