There are infinitely many test scenarios, and that is why we highly encourage to try or test a few scenes/scenarios, as close as possible to production (with a backup of course), so such things are found and tackled.
I get that the behavior is adopted, but that doesn't mean it is correct or expected, and doesn't it need to be changed or improved?
Still weird you state that a behaviour which was there since many years is "not correct" ... if it weren't correct you could have mention it before. Now that ForrestIvy is there suddenly you change things...reagarding Forrestpack behaviours. It wasn't a problem for many years, but seems to be now.
For me ForrestIvy behaves a lot like RailClone and i'm ok with railclone not having unique IDs for each subobjects, rather one object-id for the main RC-object - same with Lightmix.
But its a different scenario - i can work with this and have adapted to it.
With Forrestpack it was a cool feature you can have individual IDs/Lightmixselects per suboject and not the whole - even though in some scenarious this might be wanted. But then you could bring it as an option or figure things out with itoo... to add a option-button if masks/lightselects will be individual per suboject or for a FP-object. Then artists have more freedom with controling it.
Other than that I haven't tested it with cryptomatte yet - might be nice to have both masking options in one rendering - like whole FP object in cryptomatte and individual sub-objects in Cmasking-mask