There is another great video similar to this of Tom Dalling coding a flappy bird clone using Gosu:
By having devs that instead of following "it cannot be done", persisted and eventually were able to release something in those languages that others wanted to play.
Having watched this cycle a few times, I always look forward to see higher level programming languages used in broader contexts.
However for game development, maybe Crystal instead of Ruby would still be a better option, in terms of raw performance and distribution.
RubyMotion just badly needs a .NET/IronRuby port, or anything else that'll work well on Windows... :(
But yeah, I imagine compiling Ruby is not much dissimilar than compiling Dylan.
Actually, compiling Ruby is not that bad, horrible grammar (from the perspective of having to write a parser - as a user o the language I like it) notwithstanding.
Compiling Ruby into efficient code is a nightmare.
The problem is largely that there's a total lack of delineation of application read/load time and runtime, coupled with the ease of being able to mutate the object model.
E.g. in theory on return from any method statement the entire world might have been transformed. 2 + 2 = 4 one moment, only for 2 + 2 to have side effects and return 5 the next...
The "quick and dirty" way of compiling Ruby is to accept that you'll need expensive dynamic message call invocations all over the place.
Optimizing beyond that is a lot of work to ensure you can handle it when people start playing silly games. The jRuby Truffle backend is probably the best bet at the moment (Chris Seaton who did the initial work on that is on HN). If I ever get that far with mine, I'll be stealing lots of ideas fro there...
So what you are telling about Ruby is also an issue in Dylan, given its Lisp heritage and flexibility.