There are basically five different rendering pathways here, some of which should become more similar eventually:

• The normal autocomplete in tokenizer fields, e.g. "Subscribers".
• The @ mention autocomplete in remarkup. This has its own separate rendering because the "autocomplete" menu items don't look quite like normal typeahead items (they're a bit smaller and sleeker).
• The token which gets rendered in a tokenizer field like "Subscribers" when you use the typeahead to actually select a subscriber.
• The same token, which gets rendered in PHP and sent over the wire so that the page looks okay before JS runs, if you "Edit" an object which already has some subscribers.
• The same token, which gets re-rendered in JS based on the PHP definition after JS fires, so that it has control of the nodes and stuff like the "X" that lets you "Remove Token" can work.

There are also two other pathways:

• The more full-featured global search typeahead, which shows more information.
• The "Browse" view of possible tokens, which also shows more information.

For now, I'm not touching those.

Some day, it would be nice to get these 7 pathways down to, like, less than 7. But nothing here really makes that any harder.