Perhaps use an old Soup-bowl for some tests 😉
In editing, like in many other arts, you find a rhythm and a structure to tell a story.
In rotoscoping, one needs to find where one motion or the direction it changes to move around without too many manual adjustments.
In animation, both of the above are combined, and one or the other is used to find the "Key-Frames" that will define the main motion, which might be a bit counterintuitive, as the change of something is the key here.
Go back a few steps, and we will discuss the storyboard: which is the smallest number of images needed to illustrate the flow of the story so a team can reproduce it on set?
I assume that sounds very familiar to you.
Or, as a friend of mine coined it around 15 years ago before you start any action, try to say it out loud in one sentence what you like to do.
All of that separates each motion into sub-sequences. Each of these could have key points that stick out like a stop initially, but each individual motion could be based on a few Nulls that create a hierarchy of movements and compose the subsequence with just a few timeline entries. Those subsequences could be visible only briefly, then another rig or setup takes place and continues the motion. In other words, define each subsequence. Or Take it if that sounds more familiar.
(A rig could be a parent null that cares only about the position, the next one about the orientation, and a third one defines the rotation, etc.)
When these subsequences are done, you render a preview and take nodes.
If that flows in the way you like, put an object into sync with the positions and motions you have found.
Yes, the object you have is not a simple one to animate believably. Like any craft, it needs training and experience. Similar to that is the editing of a movie, experience and storytelling-IQ shows up heavily in the result; there is no shortcut. In other words, allow yourself some time.
My best wishes for your project