[Bug] version 2.8.0.7 Restarting broker consumes historical data

### Search before asking

- [X] I searched in the [issues](https://github.com/apache/pulsar/issues) and found nothing similar.

### Version

My brokers are using version 2.8.0.7 and there are three of them, used for handling delayed messages. Last night, due to a hardware issue, I restarted one of the brokers. As a result, all the messages that were supposed to be consumed by that broker on that day were re-delivered.

### Minimal reproduce step

restarted

### What did you expect to see?

How to resolve this issue?

### What did you see instead?

re-delivered

### Anything else?

_No response_

### Are you willing to submit a PR?

- [X] I'm willing to submit a PR!

Start by reproducing the broker restart scenario described for version 2.8.0.7 with three brokers handling delayed messages, then trace how consumed-message state is retained across a restart. Done means messages already consumed by the restarted broker are not re-delivered; the issue names no files or tests to target.

Written by the indexing model from the issue text.