Measurements do not cover the complete experiments

[mam10eks]

if you have a process that runs for multiple minutes, the measurements seem to "forget" the start.

E.g., here is an example where it starts at 251345ms.

"resources":
  "cpu":
    "energy used system": "0 J"
    "used process":
      "avg": !!int "0"
      "max": !!int "250"
      "min": !!int "0"
      "timeseries":
        "timestamps":
        - "251345ms"
        - "266408ms"
        - "275043ms"

A good behaviour would be if it reports the measurements uniformly distributed over time.

[TheMrSheldon]

The "forgetting" may appear due to the nature of the aggregation that values further in the past are aggregated more often such that the density of datapoints decreases exponentially with time. A uniform distribution would AFAIK not be possible since we want to have a fixed cap on the number of datapoints but don't know the duration of the experiment in advance such that we can't know the intervall at which to report data in advance