The main metric to debug PBTS is ProposalTimestampDifference. It produces an histogram for the observed difference from the received time for a Proposal and the Proposal timestamp. Preliminary outputs from QA experiments (#2323) show that this metric is very relevant for debugging and for defining standard values for SynchronyParams.

Two additional metrics were added in context of the previous effort to implement PBTS. They are available on main branch: QuorumPrevoteDelay and FullPrevoteDelay. They should also help the definition of SynchronyParams, in particular MessageDelay. We have to check, however, how useful are the values returned by them.

We need to check

Definition of Done

• Investigate the values produced by QuorumPrevoteDelay and FullPrevoteDelay
• Decide whether to maintain QuorumPrevoteDelay and FullPrevoteDelay metrics
• Define good representation (buckets) for ProposalTimestampDifference metric
• Define whether labeling ProposalTimestampDifference with the is_timely flag is useful
• Adapt representation (buckets) for ProposalTimestampDifference metric if default synchrony paras values are updated after the QA
• Update PBTS runbook regarding the content and use of the new metrics
• Update the metrics.md spec, align it to the new comments and semantics