telemetry_metrics.distribution is not reporting sensible results

[alco]

I am not sure whether the error is in the metric collection or in the export but, given a metric definition like

distribution("vm.monitor.long_gc.timeout", tags: [:process_type], unit: :millisecond)

I'm seeing it reported to Honeycomb (via Otel export) as follows:

vm.monitor.long_gc.timeout.avg:    668
vm.monitor.long_gc.timeout.count:  1
vm.monitor.long_gc.timeout.max:    0
vm.monitor.long_gc.timeout.min:    0
vm.monitor.long_gc.timeout.p001:   500
vm.monitor.long_gc.timeout.p01:    500
vm.monitor.long_gc.timeout.p05:    500
vm.monitor.long_gc.timeout.p10:    500
vm.monitor.long_gc.timeout.p20:    500
vm.monitor.long_gc.timeout.p25:    500
vm.monitor.long_gc.timeout.p50:    500
vm.monitor.long_gc.timeout.p75:    500
vm.monitor.long_gc.timeout.p80:    500
vm.monitor.long_gc.timeout.p90:    500
vm.monitor.long_gc.timeout.p95:    500
vm.monitor.long_gc.timeout.p99:    500
vm.monitor.long_gc.timeout.p999:   500
vm.monitor.long_gc.timeout.sum:    668

So, even though the attributes *.min and *.max are present, it's no use plotting them on a chart. I don't think the percentiles make any sense either. Since *.count is 1, there must have been a single data point collected, so the avg value is the actual value collected, most likely?

[alco]

Another example, this time the Otel metric signal contains two datapoints:

vm.monitor.long_schedule.timeout.avg:    857
vm.monitor.long_schedule.timeout.count:  2
vm.monitor.long_schedule.timeout.max:    0
vm.monitor.long_schedule.timeout.min:    0
vm.monitor.long_schedule.timeout.p001:   500
vm.monitor.long_schedule.timeout.p01:    500
vm.monitor.long_schedule.timeout.p05:    500
vm.monitor.long_schedule.timeout.p10:    500
vm.monitor.long_schedule.timeout.p20:    500
vm.monitor.long_schedule.timeout.p25:    500
vm.monitor.long_schedule.timeout.p50:    1000
vm.monitor.long_schedule.timeout.p75:    1000
vm.monitor.long_schedule.timeout.p80:    1000
vm.monitor.long_schedule.timeout.p90:    1000
vm.monitor.long_schedule.timeout.p95:    1000
vm.monitor.long_schedule.timeout.p99:    1000
vm.monitor.long_schedule.timeout.p999:   1000
vm.monitor.long_schedule.timeout.sum:    1714