Detecting stalls caused by malicious validation

[akoshelev]

We've run into several issues where validate_record was blocking protocol execution because of misalignments between batch size and active work size. It wasn't easy to detect because stall detection did not print anything on that.

We would want stall detection messages to include records waiting to be validate. Getting pending records from Batcher should be easy, but it is not clear how to integrate this with gateway's stall detection yet.