# fix\(server\): hub usage reads of one account run one at a time · pingdotgg/t3code

[View on GitCafe](https://git.cafe/pingdotgg/t3code/commit/7d6cc5b2b654f9a1d1f11be836230ff2922c4bb8)

Repository: [pingdotgg/t3code](https://git.cafe/pingdotgg/t3code)

Visibility: public

Requested revision: 7d6cc5b2b654f9a1d1f11be836230ff2922c4bb8

Requested commit: 7d6cc5b2b654f9a1d1f11be836230ff2922c4bb8

Commit: 7d6cc5b2b654f9a1d1f11be836230ff2922c4bb8

Tree: b1e5c91e56ee87a7f6ea2ae8b67fc0e2cdfbed85

Author: Yannic Charlon

Committer: Yannic Charlon

## Message

```
fix(server): hub usage reads of one account run one at a time

Two hub sources pointing at the same hub read an account at once and both
probed it. Reads of one key now take a per-key lock (the shared KeyedLock),
so a read waiting behind another gets its answer without asking. forget takes
the same lock, so it lands after any read in flight and the forget counter
goes. Entry pruning is dropped: it reset a failing key's backoff after a quiet
spell, and entries are one per hub account identity, which a restart clears.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

```

## Parents

- [54c8cf9457810a56f431c081b42369de657d56ab](https://git.cafe/pingdotgg/t3code/commit/54c8cf9457810a56f431c081b42369de657d56ab?format=markdown)

[Source at this commit](https://git.cafe/pingdotgg/t3code/tree/7d6cc5b2b654f9a1d1f11be836230ff2922c4bb8?format=markdown)