We're seeing lots(!) of data center traffic identified as Chrome on Unix in Gam logs coming from Colombus Ohio or Northern Virginia that is purporting to be in mode b treatment 1.1.

It is enough to make performance of treatment 1.1 look far worse than control 2, and also control 2 * 3 also looks far better than treatment 1.1 + 1.2 + 1.3. I am not sure if this is actionable, but perhaps a warning for others. When these traffic are removed, control 2 indeed has the worst revenue performance of the mode b groups.

I've been corresponding a bit with @rahulkooverjee-google on this and he may have additional advice.