P2P: clear inflight pooled tx requests if we don't have

[Arkenan]

After PR #6437 is merged, we add tracking for inflight transactions so we don't issue duplicated requests. However, we should check if failed/timed out transactions are cleared. If not, this can slowly accumulate memory over time, plus we will never request for those txs.

We need to check what the flow is for these timeouts, if we even have that implemented and have some level of periodic clearance, or having those timeouts clear the inflight dictionary actively when they happen.

[Avnish-07]

I think this issue is about handling inflight requests that never receive a response.

Currently, requests are removed from the requests map only when a response is received or when an error occurs. However, if a peer never responds, the request remains in the map indefinitely.

This can lead to memory growth over time, and those requests may never be retried.

My understanding is that we should handle this by introducing a timeout mechanism. We can track a timestamp for each inflight request and periodically check for requests that have exceeded a timeout duration. For such requests, we can remove them from the requests map, readd their corresponding NodeRequests back to the download_queue for retry, and mark the peer as failed.

This would ensure stale inflight requests are cleared and missing nodes can be retried properly.

Please let me know if this aligns with the intended approach

I’d like to work on implementing this timeout mechanism. Please let me know if this is okay or if there are any guidelines.