We will modify SamplingEstimator#generateSample() method so changes can be leveraged transparently by both CBR and JOO. 

We’ll need to provide a boolean flag to instruct the sampling to try reading from persistence or force fresh collection. We have to provide it internally so the analyze command can leverage it too and always force collecting a fresh collection. Other users will just feed in the value of the flag.

The idea is that at the point of beginning to collect the sample the flag is verified and if it is enabled then we construct the key according to the definition above and search in the stats collection. If found we just return that sample, otherwise we’ll fallback to the current logic.