Let me give a use case I am seeing.
I have person in my tree. I discover a source which mentions her and there is a good amount of collateral text and information about this source, so I create/define a new ‘source’. Much of this information talks to the generalities, dates, publishers and context of this source and can be put into the ‘Source Citation’ tab of the person-detail GUI, if its not already there. This dialog is labeled "Person: xxxxxx yyyyyy -Gramps. The person field also has a ‘notes’ field for general information about the person. This notes field is part of the top-level person ‘Notes’ . It is a one to many relationship between this person and these notes.
So now our person has at least one source (maybe more!) which we can cite for specific events’ instances.
For any of these arbitrary sources Gramps gives us a ‘source text’. It is also a ‘note’ object and is bound to the specific source itself. This ‘source text’ note type is presumably intended to provide general information, context etc, about this source. There is a many-to-many relationship between a person and relevant and applicable sources. One person can have many source and one source can have many persons. A citation provides a bridge from source to person.
For a citation there is the possibility of also adding a note to it. This type of note is also listed as a ‘source text’. It might be better to call it a ‘citation text’… but there it is. There is a many to one relationship between a source and a citation (there can be many citations to a specific source), and a one to many relationship between a person and a citation (one person can have many citations to a specific source).
These ontologically nested note types seem to all be Gramps notes nonetheless. Gramps stores them as different note types, where they exist, but they are all Gramps notes, it seems to me.
The note cleanup tool does not seem to address all the three categories of above Gramps notes. We should consider that there there might even be other type of notes equally un-addressed (media notes, place notes, etc).
In my case, the tool only seems to address the person notes (but does so very well!) and somehow ‘some’ of the source notes but none of the citation notes. The debris persists.
All of these notes types have the same need for clean-up since they frequently arrive via cut-and-paste of html/xml/MU/MD pages.
My question goes to whether the tool is architected to traverse all the current note types. The tool may have been developed before newer types of notes were added. Just a thought.
A separate question is whether all the possible mark up ‘debris’ in these notes is identified for the tool to act upon. Probably a harder task!