Google's the one I have the most experience with, and there's some excellent work going on there, but there's no longer a company-wide culture of excellence, and the reason is the injection of traditional HR and managerial behaviors: "calibration score" witch hunts, manager-as-SPOF, stack ranking.
If you're a Real Googler, you're above the fray with regard to a lot of that shit, and you can move to projects as you wish. Hell, Steve Yegge quit his project at Oscon and wasn't fired. He moved to a different project. The culture of excellence that exists at Google is for Real Googlers. If you're a Real Googler, it's an awesome place and you can get some really great work done there. If you're not, you're wasting your time and career. I'm sure it's wonderful if you're already great. It's not the best place for people like me who are good and trying to become great.
Valve is now the cultural leader, and it will be able to scale with open allocation. Why? Because open allocation is a natural fit for technology. I don't think you can grow at 300% per year and retain OA, but I see absolutely no upper limit on the number of people who can coexist in such a regime. What makes scaling hard is the deadline factor: having to grow quickly. It's easy if you can control the pace, which Valve (being privately funded, not a get-big-or-die VC gambit) can.
OA seems the anomaly because we're used to these military-inspired corporate megaliths, but the reality is that closed-allocation is the oddity. A caste of people whose only job is to delegate work? Top-down internal headcount limitations set according to political fights, arbitrated by people who barely understand the day-to-day of the company? Does this sound like an intrinsic necessity, or self-serving parasitism centered around control of the division of labor? I know where I'd put my money on that one.