During clearing of the card set, we currently dump the associated ConcurrentHashTable and recreate it with InitialLogTableSize.

Investigate the benefit of having a ConcurrentHashTable sizing heuristic, that is keeping the tables at some specific other size to avoid constant resizing of these.

Initial attempts at doing so showed that this is a kind of waste of space, but more investigation should be done (particular e.g. limiting keeping some particular size to young gen regions; or size them to that "good" size only before adding elements, not keeping the ConcurrentHashTable around all the time etc).