pollInterval on different pages on the same query keeps increasing instead of stopping

[chemicalkosek]

Intended outcome:

Setting up a pollInterval on a component in one page should stop that interval when going to another page (or unmounting that component) and start a pollInterval on that new page (mounted component) for the same query.

Actual outcome:
I have a page /a with ComponentA with a useQuery for QUERY. It has a pollInterval 7000. I go to page /b with ComponentB which also has a useQuery for the same QUERY. It has also a pollInterval. I then see in Network Tab that the polling frequency increases. When I navigate even further between the pages the polling frequency can increase even more, firing every second or faster.

How to reproduce the issue:

Version
I'm using react-apollo 3.1.3 so #3485 should work

System:
OS: Linux 5.0 Ubuntu 18.04.3 LTS (Bionic Beaver)
Binaries:
Node: 10.16.3 - ~/.nvm/versions/node/v10.16.3/bin/node
npm: 6.9.0 - ~/.nvm/versions/node/v10.16.3/bin/npm
Browsers:
Chrome: 79.0.3945.117
Firefox: 72.0.1
npmPackages:
apollo-boost: ^0.4.7 => 0.4.7
apollo-cache-inmemory: ^1.6.5 => 1.6.5
apollo-client: ^2.6.8 => 2.6.8
apollo-link: ^1.2.13 => 1.2.13
apollo-link-error: ^1.1.12 => 1.1.12
apollo-link-http: ^1.5.16 => 1.5.16
next-with-apollo: ^4.3.0 => 4.3.0
react-apollo: ^3.1.3 => 3.1.3

[chemicalkosek]

Actually I have just observed that even if I turned off polling on other components (pages) and left only one this issue is still there. Meaning I go to a page which doesn't have polling (but has the given query) and polling continues and starts to fire more frequently.

Could be related? #3644

[chemicalkosek]

Closing as the issue is related to when a GraphQL error exists then polling starts to increase and also continues when the component umounts (polling continues on pages where it doesn't even exist). After upgrading to Apollo CLient 3.0@beta everything works fine for now.