>>1518231my theory is that google has been experiencing an ongoing serious performance incident in their distributed message queues, and that north american data centers are hit the hardest. my theory is that there is no "spam viewer" code at all and that it's just a lie. the difference between EN and JP is not in the location of the channel; it's in the location of the audiences.
counting concurrent viewers has nothing to do with the location where the stream originates. it's a problem of collecting data from the audiences. the fact that there is *some* north american audience for holoJP is why they've been affected somewhat; the fact that there is *so much* audience for holoEN in north america is why they've been hit harder; and, the fact that, despite the NA bias, holoEN's audience is more geographically diverse than other kinds of channels is why they are an outlier who is hit the hardest.
something else i argue is that there are other things that should be affected by degraded performance in message queues. we should see dropped chat messages, and we should see the severity of it roughly track the viewer count impact. we see that. we should also see subscriber counts flap bizarrely as the count crosses reporting interval thresholds (i.e., from 999,999 to 1,000,000)... and we see that.
the other anon's independent testing agrees completely with a theory that viewers just aren't being counter correctly, even when conditions are ideal: no browser extensions, no ad blocking, youtube premium active, tab focused, sound on. for all the theories that actually rely on some kind of actual "spam viewer" implementation, there is evidence to cast doubt on it. i simply don't think there is any "spam viewer" filtering at all. their shit is broken and they are lying.
