2026.09.60 — 2026-09-29

Fixed

• Show a cluster's depth and consumers as totals, not averages (ca5be412)

  The cluster-wide Depth and Consumers charts now show the cluster's total: at each bucket's end, the sum of every queue's latest sample on every node. They previously averaged all samples in a bucket, so a cluster holding 10,000 messages across 50 queues read about 200. A queue on the slow sweep now counts in every bucket until its own sampling interval lapses, and a deleted queue stops counting after that interval. The cluster chart is labelled "Depth (total)" and has no bucket peak, since a total's peak is not the sum of its queues' peaks. Queue charts are unchanged.

• Count every increment in a counter's rate, even at a fine step (7f141864)

  A queue's rate charts (messages added, acknowledged, expired) no longer read 0 msg/s for a busy queue at the 15-minute range. Each bucket now counts what arrived since the queue's previous sample, not only the change between samples inside the bucket, so a bucket holding one sample reads the real rate and nothing added between two buckets is lost. The first bucket of a range with no earlier sample is left out instead of shown as 0. Per-node series still add up to the total.