[nightski]

While that may be somewhat true (and I disagree with the tiny fraction assessment, unless you are measuring in some other metric than time investment), I feel like it would still greatly increase your QA budget. Each additional back-end that is supported now requires extensive testing through the game development process.

[mellinoe]

In my experience, trying to get OpenGL to behave the same way (and with the same performance) on a bunch of different systems is actually more work than just maintaining multiple backends. Testing OpenGL on Windows does not in any way guarantee that it's going to work on Linux, or macOS, or whatever else -- even if you don't add a crazy OpenGL extension support matrix into the mix. So your QA budget is already going to include testing OpenGL on all those platforms.