I got 10 or so in one region, 10 or so in another region, and 32 or so in asia. It was all spread out. It really seemed like a capacity issue. This was in March

I got 10 or so in one region, 10 or so in another region, and 32 or so in asia. It was all spread out. It really seemed like a capacity issue. This was in March 2017. In fairness, Google was pretty transparent about the whole thing. It had a lot more personal touch than the other providers. Google set me up on a call with some of their engineers to discuss autoscaling. But at some point the communication broke down a bit or maybe interest was lost in helping with my issue. Concluding with the realization that they didn't support managed instance groups with GPUs yet, and that was that.