Well I'm surpised regarding games, didn't know they were firstly designed for consoles and then PCs lol. Now I understand why SLI and crossfire haven't and aren't very optimised.
And why reviewers recommend these days to get "the highest end single card you can afford" and not 2.
People call them "console ports". However, if you understand coding and how it works it really isn't a port. Games are made up of libraries, a butt load of code and wrappers etc. Then they are coded to work on a console first. Once that is done they then have to write the code so that it works on a PC. However, with recent games the work to get the game to run on a PC is much simpler because the consoles now follow the X86 format. This means that a whole ton of the code will run "out of the box" which minimises the work needed to be done to make a game run on a PC.
That is where they would have to write libraries to make the game work on your graphics card, for example. Whilst they were there it wasn't much work to code in other things like multiple core support and SLi etc. However now? hardly any of that is required so they have literally just said "Hey, CBA..".
So yeah, now it's a case of there being very little point in spending company money, time and etc adding in things like that. To a coder the quickest path to profit will always be the one they choose.
Whether or not this will change? yeah, it probably will. But, not any time soon. Now that AMD have their claws into Microsoft and Sony they will eventually release a GPU that has smaller cores in clusters. So basically multiple GPUs on one area. The reason for this is quite simple... Let me explain.
When a "wafer" of silicon is picked it's usually circular. Your dies are then basically put into that silicon wafer. Some areas of that wafer will be bad, so basically the smaller your GPU cores the higher the success rate on that wafer. The percentages of successful cores goes through the roof. This is why Nvidia are always keen to bring out their mid range silicon first. It's cheaper, and success rates are far higher. This is why they have not rushed into releasing the 1080ti, because obviously the bigger the core they need to put into a wafer the higher the chances that the cores will hit unusable areas. Thus failure rates are far higher, as is the end cost.
So what AMD have decided to do (with Navi) is basically make lots of small cores for cheap, then combine them on one "die" (a PCB ) and then get the console coders to support this "scalable" architecture.
Now when that happens (not for ages and ages, they're not even done with Vega yet) we may see "metal" multiple GPU support. IE, at code level, hard coded into the API.
Until then though? for me multi GPU is dead and not something I would touch with a 60ft pole. There is only so much Nvidia and AMD can do "after the fact" and even with games that supposedly work lately you get all sorts of flicking and texture glitching issues.
