For various reasons, the previous mechanism of allowing a new marking cycle to
commence while partially young (aka mixed) collections based on marking information from a previous
phase are still in progress was recently abandoned. See <TBD>
It would be a good idea to make the transition point from partially young (aka mixed)
collections to a new marking phase adaptive based on various factors such as heap
pressure, length of the marking cycle and efficiency of recent, current and predicted
future such collections etc. This would also seem to fit into ongoing work on G1's
(analytical) adaptive collection policy.