I think a number of us would think that the mere existence of incremental garbage collection strategies in the 80s and 90s, not to mention prior algorithmic work like heaps, ought to be prior art to kill such a patent. The important point is being able to loosen representation invariants and defer some maintenance work; this method has been applied in many systems to either increase the size of work (bigger batches) or to allow incremental work units (less jitter). (And sometimes both occur based on adaptive thresholds, e.g. batch small requests and break up bulk requests.)
The fact that overly specialized corner cases of generally know technique can be enumerated and patented is part of the problem with the current patent system. It is obvious to one skilled in the art, and it amounts to a land rush to try to lay claim before everyone else gets there. It is not innovation that needs protecting. It is disgusting.