I can address this question from pertinent experience after having spent numerous years in games before moving to film, and a big chunk of it programming real-time game graphics. Intuitively, you would think that utilizing the power and graphics of game engines for film is an obvious marriage. But after spending time and money developing my own previs tools using a licensed game engines, I learned that there is a disparity between the ideal and the realities of production. I simply was not knowledgeable of how things worked in practice at the time. And BTW, I'm really addressing professional level commercial previs, rather than independent projects.

There are many different ways that clients work with previs providers. And the cases where real-time playback matters, whether it's in maya, or motionbuilder is a relatively small portion. And here's some reasons in no particular order:

1. The number of directors/creatives that will actually sit down with you, over your shoulder, editing bay style is quite small. They are simply too busy. There are too many handlers present usually. Having said that, the most die hard clients or ones who fully embrace the previs process to make movies do work this way, but's it's a small percentage. Usually, if there is any "hands-on" time, you can quickly block really rough cameras and animation but its usually a reference for later cleanup.

2. Most previs is iterative, you create shots, cut it together, present them for feedback at a review session and get feedback, you or they go away to do stuff, and you make changes for the next round. Just getting real-time interactivity in maya doesn't offer much in solving a lot of the practical workflows in production. It helps you create shot's faster because you don't have to playblast, but you can already do this in motionbuilder, while integrating and editing mocap, and using external devices like cameras and joysticks to puppeteer animation.

3. In complex scenes, with action, fx, complex cameras, crowds, complex timing, (which is almost every previs shot), it simply takes time to animate. There are lots of keyframes to deal with and other technical appartus involved. And making changes simply takes time. It's not graphics speed that's a limitation, as we can playplast, it's the artists time to make changes to complex scenes. Nobody has the time to sit over the shoulder of an artist as they make a directorial decisions in all but the simplest of scenes.

If you think about it, this makes sense. It's common for a director to look at what you have and completely change it, in it's entirety, preserving little of the animation, or timing, or dialogue, or even location. They may say something like, "lets have this character do this instead, and he does it way over here while those guys are doing this, and let's cut this bit of action out entirely". By the time you've gotten close to what they want where the changes are small enough that you can do quickly, 95% of the work has already been refined.

4. The more visual fidelity (eye candy) you provide the artists, the more work it takes to make shots. You simply don't have the luxury in most previs jobs to finesse the graphics. You need to work fast and deliver shots to cut into an edit.

But If you start having to invest time and effort into nice lighting, effects, textures and all those those things that makes games and final shots look good, it simply does not solve the fundamental problems that previs is trying to answer...which is...how do you best tell this story. In fact, sometimes making things look too good can actually draw attention away from the storytelling, which usually is not intentional. You start getting feedback and art direction into how things look as opposed to the storytelling goal when it's not meant to address those things at all. This happens frequently!

Now having said that, sometimes making things look good, with nice lighting, detail and textures is part of the previs requirement and informs the decision making, especially when you're helping out the art department and are doing more look development. This is pretty common. And in these cases, where the visuals are a part of the problem solving, then yes, it's important.

5. Regarding mocap, it's become common to integrate mocap into previs, primarily because motionbiulder can handle multi-characters real-time, and it saves a lot of animation time while trying out different ways a scene can play out. It helps tell the story because the animation can be much less distracting if it's bad. Editing mocap is time consuming, and entire actions can and do get changed or thrown out. And frequently, there are numerous characters, interacting with one another. This is complex to change whether you are in maya or motionbuilder.

As you've pointed out, it's a pain to transfer assets between applicaitons, whethers it's between maya and motionbuilder or maya and a game engine. There's no simple solution to this because these asset tools are integrated into the pipelines of a studio the way a game developer is reliant on a game engine. Most vfx houses use Maya and so it's the primary tool for previs.

If you try to utilize new technology like game engines, you have to invest a great deal of time and resources into creating something that offers a significant practical benefit to the previs creators, that it warrants adopting a whole new toolset. It would have to be bulletproof to meet the tight schedules of previs creation, and be flexible to integrate into existing pipelines. There is as yet, anything that comes close to this. Motionbuilder is the closest thing, and is widely adopted into studios, but it's not a complete toolset and relies on other asset creation programs.

To answer your question simply, having console target visuals is nice, but only if it's incredibly robust so that it's easy to use and requires little additional artist time, than yes, that is desireable. Having real-time interactivity with full mocap tools inside Maya would be great if it was like having motionbuilder inside maya and you didn't have to transfer scenes back and forth. But they are fundamentally two different architectures with different workflows and so you pick the tool that best serves the goal. And we're still a ways off from having the perfect tools for all the ways previs is used at this point.

Don't get me wrong, we are heading in that direction, slowly but surely. And there are tools out right now that do offer some solutions. Even real-time ray tracing using GPU's is becoming a real possibility for future previs. But there still exists a great divide between the ideal and the practical realities of day to day work. No single tool or technology is making the process significantly easier YET. It's evolutionary rather than revolutionary. Creating quality shots that makes clients happy just takes time effort and great artists and storytellers are your best guarantee.