fix: re-evaluate derived props during teardown

[Rich-Harris]

alternative to #16263, made possible by #16270. Closes #16114 (but not #16262 at present, investigating)

Before submitting the PR, please make sure you do the following

• It's really useful if your PR references an issue where it is discussed ahead of time. In many cases, features are absent for a reason. For large changes, please create an RFC: https://github.com/sveltejs/rfcs
• Prefix your PR title with feat:, fix:, chore:, or docs:.
• This message body should clearly illustrate what problems it solves.
• Ideally, include a test that fails without this PR but passes with it.
• If this PR changes code within packages/svelte/src, add a changeset (npx changeset).

Tests and linting

• Run the tests with pnpm test and lint the project with pnpm lint

[changeset-bot]

🦋 Changeset detected

Latest commit: 99f0464

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
svelte	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[svelte-docs-bot]

Preview: https://svelte-dev-git-preview-svelte-16278-svelte.vercel.app/

[github-actions]

Playground

pnpm add https://pkg.pr.new/svelte@16278

[dummdidumm]

Even if there's a fix to be found for #16262 using the existing mechanism it won't fix the similar but different issue in #16072. I think we really need to widen the "keep last value on teardown" logic for props, or even make them their own signals, i.e. not pass them through but instead create "barriers" so the old value stays around.

[dummdidumm]

what's the perf impact of these changes?

[Rich-Harris]

Negligible to non-existent — the only time there's any extra work at all is when reading deriveds in effect teardowns, which is rare, and it's only a tiny bit of extra work. pnpm bench:compare corroborates this