There are some API methods that result in a response containg values that are different each time the request is sent (a good example is the 'Random' API method from the rippled server). To not let this influence the clustering, it is good to know which values are 'random' so they can be left out of the feature vector which is used to cluster responses.
One way to do this, is to sent the request twice (with a reset server) and check which values are different. This does however cause some overhead (slowing the program down to twice as much). For now however I do not have a better method which is just as reliable so I will implement this and make it an option (whether or not to send requests twice).
Some tweaking could also be done to not have to send every request twice.
There are some API methods that result in a response containg values that are different each time the request is sent (a good example is the 'Random' API method from the rippled server). To not let this influence the clustering, it is good to know which values are 'random' so they can be left out of the feature vector which is used to cluster responses.