In WT-6709 replaced the timestamp queues with a new mechanism that involves scanning a list of sessions.

To monitor this we also added a metric txn_walk_sessions that tracks the number of times that we scan the sessions. However the work done depends on the number of sessions scanned each time. Would it make sense to also add a metric, e.g. txn_sessions_walked, that records the total number of sessions scanned, so we can easily correlate this mechanism with any performance issues?