There are many ways the script could do cache busting, but it seems sometimes it actually works against our best wishes. Real solution to this should be investigated, and if there's a way to fix it it should be added into the .json so that after loading it for the first time the script knows how to refresh it correctly.
This could also just include an incremental version number i.e filename part of the path is the cache bust not the uri query. For example templateManager is told to add https://littleendu.xyz/template.json
as a source, and then in the .json templateManager would find "hey, we actually have updated this 45 times. You can find the templates in https://littleendu.xyz/template-v45.json
and in the future, try https://littleendu.xyz/template-v46.json
to see if we have an update"