It would be great if HoverImage could work not only when hovering over a note, but also when hovering over a word within a note
Are you talking about the note’s $HoverImage, or something different,? IOW, defining image-containing tooltips for words/sequences within $Text? If so, how do you envisage these anchors being set?
Out of interest, what is your envisaged use case for this?
For example, do not place figures in the explanatory text, but only when you put the cursor on the name of the figure will pop up the figure itself
You can link the right picture to the right word by analogy with footnotes
[to be clear, writing as just another user]
Footnotes, within Tinderbox, are implemented as a link (type). I can see the same (link) mechanism potentially supporting a tooltip (i.e. an image in a pop-over). The only challenge is where link anchors overlap. Although image roll-overs don’t (shouldn’t!) need a click, there is he edge case of an unthinking user defining overlapping image anchors.
Were the above to be implemented, my suggestion for the edge case is the first (by run-of-text order) takes precedence. IOW, any overlap ‘belongs’ to the first image link. If that doesn’t suit the user, it’s up to them to re-define the links or use better text!
In truth, I see a wider use here as short-form annotation, i.e. showing text. the challenge becomes getting users disciplined enough to not make a 10k word pop-up. 
Point being, there will always be—to that user—a valid reason, yet that reason often gives no acknowledgement to the realistic constraints of the implementation(e.g the engineering cost of making a scrolling pop-up text if text is too ling). My experience (from commissioning) is UI is often much more expensive to engineer than imagined. IOW, the tyranny of 
fixed length tooltip is a practical accommodation.