CA-411679: Runstate metrics return data over 100%

[BengangY]

To handle deviations in CPU rates, Derive values exceeding the maximum by up to 5% are capped at the maximum; others are marked as unknown. This logic is specific to Derive data sources because they represent rates derived from differences over time, which can occasionally exceed expected bounds due to measurement inaccuracies.

[last-genius]

While the bulk of the inaccuracies should have already been addressed, there can potentially still be some small measurement inaccuracies (especially at high load) or floating-point related errors. 5% does seem more or less proper as a cutoff point for these.

Maybe we could, at least for some time, log something every time such a cutoff occurs?

[BengangY]

While the bulk of the inaccuracies should have already been addressed, there can potentially still be some small measurement inaccuracies (especially at high load) or floating-point related errors. 5% does seem more or less proper as a cutoff point for these.

Maybe we could, at least for some time, log something every time such a cutoff occurs?

Good idea. I saw there is stats.ml in rrd lib for timing statistics, but it seems not used much. Maybe we can use it to do some statistics for these inaccuracies.