Would game developers really want to make games for a VM? Don't most big games usually try to squeeze as much out of the hardware as possible? (Genuinely curious... I know next to nothing about the gaming world.)
Game developers are fundamentally human. They care about people playing games, nothing more nothing less. You don't need to necessarily make your game have the latest and greatest graphical glitz. If they can make their game run smoothly on your hardware, they won't care if it's in a VM inside a Virtual Box on a cloud server, under the sea.
Minecraft is written in Java. There were games written in (at least partially) Haskell/Erlang/Ruby/Lisp. Unity3D is based on C#, etc. If tradeoffs are worth it, they'll do it.
While developers surely are, game development companies are not fundamentally human. Now that Bobby Kotick has finished running Blizzard and the FPS genre into the ground, Valve is pretty much the only major publisher left that hasn't been overrun by IP lawyers and MBAs. Minecraft is a great game, but it's not a good representative of the average successful game. Developers make games because they want people to play them, but unfortunately, the only ones who have any control over what they are doing are the indie devs.
We already have indie / casual games on Linux. That's not enough to make Linux a legitimate platform.
That usually only applies to performance-intensive code like rendering, physics, pathfinding, etc.
The actual gameplay is often implemented using scripting. For example, with Unity you write games in C# (with Mono) using their engine, and it's still fast enough to work on cellphones.
There used to be a time when game developers didn't use "high-level" languages like C++ at all. But in the end they moved to them because of other benefits they got, and most of them slowly started to realize that with the hardware getting faster and faster, it's a good trade-off to make. They might think that again.
You are correct. The mainstream gaming industry would have absolutely no interest in developing for a VM, because developing games is kind of like a nuclear arms race.
Unless someone finds a way to get Next-Gen 3d Engines to run in a VM without a performance hit of even 1%, its not going to happen.
And what would the point be? Massive loss of performance by going through a VM - only to save 30 bucks on licence cost for e.g. Windows? Any game dev wants the most performance they can get their hands on.
If the gatekeeper (apple/ms/whatever) decides to tighten the screws, taking not 30% of your revenue, but, say, 70% revenue - then fsck the performance, it may be quite reasonable to do major changes in art, gameplay and coding practices just to fit to a more lucrative market model.
And, as is well known, simply the existence and theoretical threat of a plan B (in Gabe's case, Linux) ensures a much better negotiation position and a better (more fair?) plan A.
I'm not sure if you have used that to do any kind of intensive code. I've used it for real time video processing (H264 encoding/decoding and playing from a custom protocol) and it's not a good development experience in the least. It's extremely hard to get good debugging working, eclipse is a giant pile, most of the other IDE's for Android don't support C/C++ ndk, and there a mountain of little special compiler tricks and options to learn. Not to mention the JNI tricks and hidden issues. It's far from ideal. I REALLY!!! wish google would step up and provide a real solution to the problem there. Compared to how easy the same things is on iOS, Windows mobile 8, or bb10 is fairly frustrating.
