I added proper documentation to the function, adjusted the function name to be more global and I did use the helpers function doge#helpers#count() to simplify the code. I moved your function code to the TextChangedI callback function and there it will properly update the new todo_count variable in b:doge_interactive. Your function will then only check if the amount of TODO items is 0 and will deactivate when the statement evaluates.

Any suggestions / recommendations or possible drawbacks towards my implementation I haven't thought about?

Unfortunately this way TextChanged doesn't work, because if placeholders are deleted in normal mode that counter won't be updated (because it's updated when jumping). I tried a similar approach in my very first PR (checking the count) and I didn't like it. If you want it to work with TextChanged I don't think there are alternatives to search for the pattern in the autocommand, but it's not that there are performance issues or anything, so I don't see the problem.

Agree, I haven't thought about that, but wasn't for of this while I implemented it. I changed the implementation by moving the update part in the TextChangedI function to your function. This works perfectly as I see for now. Agree?

You sure it's working now? tabbing in insert mode seems broken to me.

I do have the same. Didn't test insert-mode last time. This has to do with the doge#helpers#count function. It seems to be the let l:cnt = execute(l:range . 's/' . a:word . '//gn') line and I have no idea why. If I remove this and just return 4 then it doesn't occur that the count returns 0.

EDIT: This is only at the InsertLeave auto-command. Removing autocmd insertleave * call doge#comment#deactivate_when_done() immediately resolves the problem.

I think you should learn a thing or two from this guy. Then maybe you can tell me what was wrong with the function I wrote :).

I reverted back to your previous commit. The reason I prefer my version is because your code is redundant while there is a helper function that should work, but the helper function is not being called at the auto-commandInsertLeave so I'll just leave it at your version and I will do more research about this behavior this weekend. For now your version is fine.