FWIW I’ve been working on a related solution for Antora, where the index table and components are on different pages.  In that case the role of the tree processor is taken by querying the Antora content catalog, which can be done in the block processor.  The (javascript) code is at David Jencks / asciidoctor-antora-indexer and I’m working on using it for Camel at CAMEL-14874Generate component index page tables with Antora

Since my solution is intended to be fairly generic I do not have the block macro generate the entire table but rather run only when it immediately follows an existing table, to which it appends rows.  That way the block macro doesn’t need to deal with the complexity of generating the entire table structure, formatting, columns, header, etc etc.  If you adopt that strategy I think your block macro could basically serve to locate the table to append to, and then the tree processor actually do the appending.

David Jencks