Can observable sharing be achieved in a pull based FRP model?

[clinuxrulz]

The question is pretty much the same as the title.

Can observable sharing (work sharing) be achieved in a pull based FRP model?

What optimisation exist for pull based models?

Why do I ask?, Pull based models offer simpler implementations for back ends that support GC but do not support weak references. As they will not required the developer to implement a garbage collector in order to work in such languages. Yes I'm looking at you JavaScript. But I do not indent to write such a library, and I do not intend to replace sodium-typescript (its a great library), its just more for interest sake.

I know pull-based is considered slow, but being able to dodge the need for implementing a GC is rather attractive.

[clinuxrulz]

I think I may of asked a silly question. The answer seems to be as simple as just a cache per node that is invalidated if any source nodes' caches are invalidated upon value request.

Edit: I've been using the typescript version for a while, and I've realized my memory management mistakes I have made along the way, and its awesome. There is no way I'd try to rewrite the typescript version, it is perfect the way it is.