bug: filter subscription stops without occasional pings

[vpavlin]

This is a support request

Problem

I created waku-dispatcher some time ago to overcome some of the drawbacks of js-waku - mainly the unreliability in peer connections and message delivery.

I tried to update to latest version (0.0.23), which solved my issues with lightpush (and not being able to publish), but it is somehow breaking filter and my example app fails to receive any messages.

Let me describe the flow:

1. Dispatcher gets initialized
2. dispatcher.start() then subscribes to the configured contentTopic
3. Dispatcher sets up an interval where it attempts to check the subscription (unsing ping) and it tries to unsubscribe and subscribe again if the ping fails

When I publish the message using the example (https://github.com/vpavlin/waku-dispatcher/tree/chore/update-waku/example) it initially gets delivered fine, but over time the delivery (using filter) starts failing and then even the re-subscription attempts start to fail.

Various errors seen:

One way to overcome the errors is to clear the waku:peers localStorage field - everything works fine for ~5-10 minutes after I clean it, but then the errors come back again.

It is highly possible, that my subscription or peers handling is wrong in waku-dispatcher, but I cannot find any obvious issue

How to reproduce

git clone https://github.com/vpavlin/waku-dispatcher.git
cd waku-dispatcher
git checkout chore/update-waku
npm install
npm run build
cd example
npm install ../
npm start

Notes

[vpavlin]

I am running https://github.com/waku-org/lab.waku.org/tree/master/examples/light-js with updated waku/sdk for over an hour, so the problem above probably stems from my attempts to overcome the unreliability:D

[vpavlin]

Actually, that failed after some time also. The strange part is that everythign works perfectly for a while when I clean up the waku:peers localstorage cache.

[weboko]

The task is a bit outdated. Current behavior in the latest version(waku-org/lab.waku.org#60) is as follows (according to @vpavlin):

• when tab is left unattended for 5 minutes - Filter stops to receive messages.

It is consistently repro for me and easily fixable by using periodic pings.

Problem is addressed in #1970

Next steps:

• check how it works if tab is suspended by the browser;
• check if connection is kept after 10+ minutes;
• in case all peers are disconnected (currently Filter keeps 3) - populate them again automatically;

[vpavlin]

Should I test the PR somehow? HAppy to help verifying it fixes the issues