Silver figured out that this was the cause of pinging out on IRC when using ChatZilla in the sidebar (tested with chrome://chatzilla/content/chatzilla.xul bookmarked and "Load this bookmark in the sidebar" ticked, then unfocus sidebar and idle for 10~20 minutes). Because content in the sidebar is always visible, any scripts being slowed down like this would be highly noticeable. The sidebar should be excluded from the timeout clamp. The workaround of setting dom.min_background_timeout_value lower isn't acceptable as it's not fine-grained to deal with the sidebar case so defeats the benefits of background tabs being clamped.

Why is the sidebar marked as an inactive docshell? That's your real issue; the background clamp is only applied to inactive docshells, and anything the user sees should NOT be marked as inactive (because we already throttle layout updates in those, and are about to work on not doing any invalidation or painting either).

And by the way, I can't reproduce so far. I added a bookmark for https://bugzilla.mozilla.org/attachment.cgi?id=522433 to load in the sidebar, and it's not being clamped to 1s....

Ok, looks like "any scripts" was wrong, since I also see the ping timeout notification happen in real time. Then the question becomes why does setting dom.min_background_timeout_value to a lower number (64 in my case, nostalgia ;-) result in no more ping timeouts?

You tell me. If you follow your steps to reproduce and breakpoint in the global window code on a setTimeout being clamped to the background timeout, do you hit the breakpoint? If so, what's setting up that timeout and in what window?

Nevermind, after posting my last comment ChatZilla is still pinging out even when I set that value to 5. So I guess Silver was wrong and this is invalid.