Currently persistent samples have a different default sampling method to OTF sampling. Random is preferred for persistent samples due to the improved sample quality, while chunk is preferred for OTF sampling due to planning performance.

However, this may introduce a problem when persistence of samples is automated. All samples that are persisted need to be random samples so that they will be picked up on the read path.

Another thing to consider is that some collections will continue to sample OTF every time once persistent sample storage is used up, so we need to take into account the impact of random sampling OTF on planning performance in those cases.