I don't use it anymore though because I found the suggestions to be annoying and distracting most of the time and got tired of hitting escape

Same. It took longer for me to parse and validate the suggestion as it did for me to just type what I wanted.

I do like the helper for more complex refractors.
Where you have a bunch of similar, but not exactly the same, changes to make.
Where a search & replace refactor isn't enough.
It manages to figure out what you are doing, highlights the next instance of it and suggests the replacement.
I don't think I've seen it make a mistake doing that, and it is a useful speedup.
I guess the LLM already has all the context: the needle, the haystack and the term.