With CPG and Kyle's chrome->content snipping patch, I wonder how often will need the cycle collector to tear down a page, for instance when you close a tab. It may be reasonably common that after you close the page you end up with a DOM and a bunch of compartments that form a closed subheap. If we can somehow detect this, then we could free the DOM without requiring that the CC trace things all over the place. On the other hand, there may be enough other ref counted things that are associated with the page that aren't part of the DOM that this isn't feasible.

A sort of similar idea would be to treat an entire compartment as a single object in the CC, at least for tear-down scenarios. Then you'd avoid having to trace through the compartment. Though you'd still have to examine every object in the compartment to find nsISupports pointers, but maybe that could be done more quickly.

Is this a WONTFIX now that bug 754495 landed?

No. At least we could improve content blasting (bug 730639) to handle also documents (That can be a bit tricky since documents tend to be owned by many different things).

This is a little vague, so I'm just going to close it.