Browser tests often fail with "[no http server activity]"

[sbc100]

I believe what is happening here is the browser dies or becomes unresponsive, causing all future tests to fail in the same way.

I don't believe our current retry mechanism helps in this case because it doesn't do anything to restart the browser.

I think we need some way to kill and restart the browser in this case.

See #20584 and #20531

[sbc100]

@kripken WDTY

[kripken]

If the entire browser crashes or is unresponsive then restarting it is the right thing, yeah. And that could be the reason. Another possibility is that only the browser tab crashes ("Aw, snap"), in which case opening another browser tab (without restarting the browser) might be enough. But I don't have any idea either way, and a full restart seems simpler, so sgtm.

[sbc100]

As long as we are using our current test harness mode, I don't think we have a way to kill just one tab.