It should thrown an exception and abort the script with an ugly error traceback.
If the documentation says "float" and you can hand it a numerical string, then either the function or its documentation have a bug that needs correcting.
Asking a function that formats a number and that is defined that way in the documentation to format a string is a programming mistake. As the discussion clearly states, it's a minor change in application code that would make it both clearer and more robust.
I would go as far as recommend using a USB shock device to issue a warning to the programmer every times a mistake like this is made. ;-)
It should thrown an exception and abort the script with an ugly error traceback.
If the documentation says "float" and you can hand it a numerical string, then either the function or its documentation have a bug that needs correcting.