There are a few cases where the sizeEstimates that we receive from Cassandra can be extremely inaccurate and lead to a sever overestimation of the data in Cassandra. This is mostly only possible when not using vnodes or if there is an extremely small token range with a very huge partition in it.

We will now warn and fall back to our minimum partition threshold if we find that our data size estimate is Long.MaxValue (sign that our estimate was too large)

Note once we start dealing with
9.22337204 Exo bytes sized clusters this will need to be fixed again...