> The particular case is this:  I have objects (tasklets) that depending 
> on their runtime state, need to run a finalizer or not.  If not, they
> can be collected.  If they do, then they end up on gc.garbage.

But why don't you try using tp_dealloc as I suggested instead? It will allow you to run a finalizer in all cases, without anything ending up in gc.garbage. It's what we do in the new io lib.

Also, I'm still worried that the mechanism you're proposing is a bit quirky. I hope other developers can give opinions or even suggestions.

> But since these patches of mine seem to be met with rather limited
> enthusiasm,  I wonder if I should be bothering.

Well, the GC is a rather delicate affair and moreover it doesn't have a dedicated maintainer. That doesn't mean your patches are not interesting, but I think everyone tries to be very careful with that area of the code.