Kévin Le Gouguec writes: > Can anyone confirm that this would (in principle) be the way forward, or > tell me if I am missing something[3]? I went ahead and cooked up a proof-of-concept patch, which (1) adds safe-local-variable properties to org-todo-keywords and org-todo-keyword-faces, (2) stops applying default-value to org-todo-keywords, (3) delays regexps/font-lock setups until after file- and dir-local variables have been set. While this patch contains a few things that make me weary[1], it solves my use-case, and passes the current test suite with Emacs 26.3 and 28. Does this look sound overall? Does anyone have any idea what kind of breakage might be slipping through the test suite? Thank you for your time. [1] - It's hard to feel confident that moving org-regexps-and-options and org-set-font-lock-defaults like this will not introduce regressions. - Also, these safe-local-variable validation functions could probably use some unit tests.