https://dplyr.tidyverse.org/reference/dplyr_extending.html

---

If you have data frame attributes that don't depend on the rows or columns (and should unconditionally be preserved), you don't need to do anything.

If you extend a "data.frame" rather than a tibble, then you need to provide a [.class method to make it preserve attributes. See #6689 and #6691

---

Current usage section for mutate() should be updated to also include that it uses 1-D [ to implement .keep, and that is unconditionally called even if keeping everything.

---

Maybe do a pass through everything in Current usage to see if we are up to date