[vitest-pool-workers] Reintroduce isolated storage

[NonSpicyBurrito]

Describe the solution

Isolated storage was removed as part of migration to Vitest 4, however:

• Isolated storage was a well documented feature with no hint of it being temporary/deprecated/to be removed in the future.
• Running each test in a fresh environment is a common (many consider it as best) practice in backend development.
• There is no migration path provided, you are forced to simply rewrite your tests to not rely on isolated storage.

This effectively blocks many large projects from upgrading without significant effort.

I would like to request reintroducing isolated storage, either exposed as a config option or as a public API that we can manually call in beforeEach/afterEach.

[NonSpicyBurrito]

In my project I have ~4000 integration tests across ~300 files, the current lack of isolated storage blocks migration.

[penalosa]

We've removed isolated storage as part of preparing for v1 of @cloudflare/vitest-pool-workers, and will be updating the documentation soon (#28908). It's worth noting that you can still have per-test-file isolation, it's just the per-test-block aspect of isolated storage that's been removed. This leads to a lot of stability improvements, and cleans up a long tail of outstanding issues we'd seen with Vitest, that were all due to isolated storage.

There are two recommended paths forward:

• Write your tests that need isolation in separate files, which will have isolated storage.
• Add a beforeEach()/afterEach() that cleans up the specific data you need cleaned up in every test file (using e.g. env.KV.delete()). This maps to how other platforms and database offerings work, and is a more explicit way of marking the specific data you care about cleaning up.

[penalosa]

Reopening to track adding a deleteAll() helper to delete all data from bindings.

[MattIPv4]

👋 @penalosa, with this now being a discussion, is this still the right place to follow updates on this helper being implemented, or is there a separate issue (or PR) tracking this?

cc @petebacondarwin

[penalosa]

We're still tracking it here. The relevant workerd API was just merged today (import {deleteAllDurableObjects} from "workerd:unsafe"), and so we should be able to introduce this helper very soon.