Ingo Schwarze writes:
>  * One needs to consider whether using ???locale(3) in the shell
>    carries any risk, or needs to be restricted to certain areas of
>    the code.  It's not yet clear to me that suddenly running all
>    the code in the shell under a locale different from the C locale
>    would have no detrimental consequences.
>  * Security and reliability implications would have to be considered.
>    For example, in some situations it might be a catastrophic
>    vulnerability if the shell would somehow end up dying from EILSEQ.

For what it's worth I strongly suspect that the existing double-width
handling in ksh causes this crash:

https://marc.info/?l=openbsd-bugs&m=172993414312117&w=2