Endless failed request polling loop

[Gudahtt]

The polling block tracker has been observed being stuck in an endless loop (e.g. MetaMask/metamask-extension#17040).

I was able to reproduce this scenario by using MetaMask extension to switch to a network that will fail to respond to requests. This results in an endless loop of failed requests, even after switching away (i.e. even after there are no more listeners). In the debugger I confirmed that the only active listeners were internal to the block tracker itself (specifically this listener).

[Gudahtt]

I was still able to reproduce this bug on v9.0.3

[mcmire]

In the PR that is linked to in the timeline above and was closed, I made this comment:

@Gudahtt and I had a discussion and we feel like we can solve this a different way that doesn't involve listening to latest in any form.

@Gudahtt Do you still think following that route would be a proper fix here?

[Gudahtt]

Unfortunately I don't recall the details of that solution

[mcmire]

For clarity, a solution here would be to figure out a way to have the block tracker ignore internally-added listeners. More details here: https://consensys.slack.com/archives/C06FXU326RL/p1718910083555919?thread_ts=1718130645.943409&cid=C06FXU326RL

[Gudahtt]

^ Since the Slack thread is gone now, I will summarize from memory:

The block tracker keeps running as long as there is a listener attached. But unfortunately it's keeping itself alive in a loop because it's adding a listener to itself internally. We can fix this by disregarding internal listeners somehow when deciding whether to keep the block tracker alive or not.