It may be a required optimization to cache results of those method calls which analyze classes for containment of nodes like annotated fields and methods like DeconstructPattern#takeExtractMethods(V value).
Guava Cache seems to be the best implementation fitting these requirement.