Memory leak: devtools holds reference to "cleanup" memo

[aquaductape]

Solid devtools created_owners inside s.DEV.hooks.afterCreateOwner is holding reference to createMemo (from Child component) inside a Show that is supposed to be cleaned up.

The leak doesn't happen if chrome extension is connected.

const Parent = () => {
  const [value, setValue] = createSignal(1);

  return (
    <>
      <button onClick={() => setValue((prev) => prev + 1)}>Refresh</button>
      <br />
      <Show when={value()} keyed>
        {(_) => {
          return <Child />;
        }}
      </Show>
    </>
  );
};
const Child = () => {
  const allData = createMemo(
    () => {
      return new Array(1e5 * 100).fill(Math.random());
    },
    undefined,
    { name: "ALL_DATA" }
  );

  return <div>{allData()[0]}</div>;
};

Steps to reproduce:

1. Download repo: https://github.com/aquaductape/solid-devtools-memory-leak
2. Open Memory tab in chrome devtools
3. Select "Allocations on Timeline" and then press "Record" icon
4. Click Refresh button to fire click
5. Keep clicking at least 10 times
6. Stop recording on Memory
7. Look for "{fn, owned, sources, sourceSlots, cleanups, value, owner, context, pure, name, observers, observerSlots, comparator}" or just search "name" on filter
8. You'll find the duplicated memos that is retained by SolidDevtools with name: "ALL_DATA" based on cleanup amount, how much you triggered via click.

Also you can check Chrome Task Manager or devtools Performance Monitor to see JavaScript Memory increase

[thetarnav]

Yeah, I was aware of that when I wrote the setup logic.
I wasn't sure how to exactly solve this.
It's either to have a few-second timeout that will clear the array, and so the extension will have limited time to connect and use these owners.
Or also use a FinalizationRegistry, just like for unowned signals. I don't know why I didn't just use that. There might have been a reason that I'm not remembering now.
Or tap into the owner's lifecycle, to know when to remove it. Which again there might have been a reason for not doing that.
I will need to analyze this again and come up with something.
Thanks for making the issue though, I wasn't aware that it's actually so noticeable.