When geoprocessing tiled annotation, the result can be automatically display on the content like grouped by scale :
A step further could be to do the same grouping in the database itself, maybe by generating Feature datasets for each scale and naming them with this scale, to make maintainability easier. If something gets wrong or we lost this grouping is pain in a ask to select each layer to get them back on a map or deleting them if no more use of it.