Bulgaria, Europe here. Have some more technical info to provide.
During mornings (HQ nights) API timings vary between 200-400ms. But later in the day as America starts to wake up latency starts spiking with increasing frequency and google cloud (the dataset, cube segmentation and OV mesh host) starts slowing down. I have seen calls like /1.0/task/123456/aggregate and /1.0/task/123456/neighbours take upwards of 4-5 seconds to complete (they’re called when you click a cube in the OV).
During the marathon today I even saw the scythe heatmap taking 15 seconds (!) to load. I think such API timings are unacceptable and should be resolved to ensure a smooth experience.
Example: