Starting about 5/24 or 5/25, we see a massive increase in the number of watch establishments from all clients to the kube-apiserver during non-upgrade jobs. While this could theoretically be every single client merged a bug on the same day, the more likely explanation is that the kube update is exposed or produced some kind of a bug.

This is a clear regression and it is only present on 4.17, not 4.16.  It is present across all platforms, though I've selected AWS for links and screenshots.

4.17 graph - shows the change

4.16 graph - shows no change

slack thread if there are questions

courtesy screen shot