Seems Yahoo! tried to solve this with their "brickhouse" program, a new product incubator, but ultimately it seems that was too disruptive and got axed.
If you put some set of people into R&D roles where they have full autonomy, then everyone wants it. The "Real Googler" issue (~10% of engineers have 20%T and the freedom to move around the company; the rest have manager-as-SPOF bullshit) is definitely a source of resentment at the Big G.
If you're going to do that, then have open allocation for all engineers. It's a much ballsier step, but it avoids resentment of the "special friends club" that gets to do cool R&D and that the rest of the company ends up undermining.
The value in a incubator, new product development group, or "spin out", (don't think any CoE I've seen fits this model...), is that they have top level sponsorship to go outside of normal process for: hiring, sales, tools, platforms, etc. as required, and just build new stuff. If that's disruptive in the short term, C-levels should be pushing it through for long term gains.
You also place a lot of trust in engineers to have Clue about what to work on. Most will have ideas, many of those ideas would be a very poor use of the companies resources. If those guys feel strongly, they should spin out, and pitch to their old senior execs for seed funding. They'll soon find out if their idea has legs...