Put it off as long as needed to get K1 in there. The quicker we get Kepler+ (or any desktop based gpu) into mobile the better. More games will be made in OpenGL to address mobile & PC (and console if they live past this year or next as Socs hit 20nm etc) and that saves devs time/money. The months shaved off the port jobs gives extra profits or extra time for working on the game itself instead of the port process.
2yrs until DX12 (well, 1.75), I hope OpenGL is dominant by then. Once people are use to it DX12 will be a tough sell. You would stay OPEN rather than go back to Dx12/microsoft if you'd been using it for the last 2yrs. DX12 would have to have something you want that OpenGL 4.4 (probably a 5.0 etc by xmas 2015) etc can't do. MS would just have to write an OpenGL driver for the xbox1...LOL. We know the hardware can do it.
Ha Mircrosoft would rather dive balls first into a vat of acid then do anything to help the market if the market is competition. OpenGL and the Kronos group need to keep up with technology. Right now the developers are the slow ones, not the API's. Tescellation really hasn't even had it's day yet. However, if Android TV is what Android OS is, then we can expect crap (java) on top of Linux. If it gives developers access the Linux kernel, there might be hope to this.
If they really want this to be successful they need to give a TV experience plus, instead of just the plus. Your cable box doesn't "boot" to the TV Guide, it turns on to a channel so users can immediately watch TV if they chose to. They need to find someway to overlay on top of traditional programming via HDMI pass through, or store "favorite" content locally to begin playing as soon as the device detects its input is active.