Investigate hypothetical autotracking-based perf optimization

[markerikson]

https://github.com/simonihmig/tracked-redux uses autotracking to selectively update UI components that depend on specific changed pieces of state.

I already ported that logic to Reselect, in reduxjs/reselect#605 .

I wonder if there'd be some way to use this within React-Redux.

Like, what if we had an alternate Subscription implementation that wrapped store.getState() with this autotracking proxy and passed that around, then tracked which callbacks accessed the state and got smarter about when to re-run those?

I'm sure there's a ton of caveats here, such as nested subscriptions and connect, bundle size, perf overhead, etc.

related, yet another "Redux perf is hard at scale" talk: https://twitter.com/zeigenvector/status/1668713918049525760

[guillaumebrunerie]

Just saw this issue now :) Please take a look at #2043 where I describe a potential way to use SolidJS's stores to avoid rerunning irrelevant selectors.

[markerikson]

There's some strongly-related work from the Causal team here, where they've built their own store subscriptions and are trying to narrow down how many selectors run based on ordering:

• https://twitter.com/lukaskoebis/status/1721535684748464494
• https://www.causal.app/blog/re-re-reselect