Description

In the current blobs I am using tokio::sync::watch for notifying the observers when the state for a hash has changed. This works, but is way more heavy than needed.

1. tokio::sync::watch is optimized for a case where we have tons of watchers. It uses a giant BigNotify under the hood to solve the thundering herd problem. I think it will be much more common that you have a few tasks operating on a hash (frequently just one), and a few observers (frequently none).
2. since we now know that all tasks for a hash run concurrently but not parallel, we can use the awesome AtomicRefCell instead of a full blown lock.

Breaking Changes

None

Notes & open questions

Change checklist

• Self-review.
• Documentation updates following thestyle guide, if relevant.
• Tests if relevant.
• All breaking changes documented.