Timeline
The API does provide access to the timeline and to the action (feature, construction geometry, sketch, etc.) that resulted in that timeline entry being created. Each entry in the timeline is represented by a TimelineObject and it supports a healthState property that returns some information about the current state of that object. For example, you can find out if a feature has failed. By using this and the ability to access the feature associated with the timeline object you can get additional information that _might_ enable you to infer some of the information you listed. In theory it sounds like a good idea to remove unneeded features from the timeline but in practice it's not so simple. Things might appear not to contribute to the end result but in fact do play a part in the recompute.
For example you specified that an occurrence might be grounded and then later ungrounded so the net effect is that it's ungrounded. That's true when you're looking at the end state of the model, but if you roll back the timeline to a point in between those two nodes it will be grounded and there could be some action in that section of the timeline that depends on it being grounded. Deleting something that seems to have no effect could cause other unwanted side effects.
I suspect there are cases where you can have operations that can be safely deleted but it's not obvious to me what those would be and how you would detect them. If you have some known cases you can play with the API to see if you can get enough information to identify them.
