I agree that the example by @WUlate nicely illustrates the complexities that may arise for the relations between collections in virtue of the histories of their constituent specimens.
Handling old collection codes as “ambiguous synonyms for multiple new collection codes” might fall short of achieving a meaningful representation of the actual relations among the collections concerned.
I would recommend drafting a carefully chosen and defined set of relations between collections that corresponds to user expectations in the context of the catalogue and that represents collections’ relations along the lines of splits, merges, and in general transformations in terms of their constituent specimens (and I’d gladly try to contribute to such an approach).
This could be done so that both relations among collections existing in a temporal sequence (e.g., a historical collection being absorbed wholly or partially in a contemporary collection) and among collections existing concurrently (e.g., some specimens transferred from one to another in workflows such as the one described by @WUlate).
Last but not least such an approach could also support cases where collections, even concurrently, are conceived in terms of different partitions of the attribute space (“dimensions”) of an institution’s (or several institutions’) holdings, e.g. one partition based on taxonomy vs. one based on preparation type.