JT has rewritten the JSON and search page generation code so that filter fieldsets on the search page now have ids which match their JSON files; that means the JS can trickle-retrieve them after startup, allowing for the normal sequence to be superceded if a search is initiated during the process. I have the JS retrieving them successfully; now I just have to tweak the search-preparation code so it can get specific ones more quickly if required, and then tweak the search execution code to use the new split-out filters instead of the old docs.json. After that, we can dispense with docs.json. 120 minutes.