Exactly as I said in the immediately following part of that sentence: Unicode characters with codepoints greater than 255.

The string "\x{61}\x{1f60e}" has a length of two, it consists of two characters. Its internal representation happens to consist of 5 octets, but any time you have to care about the internal representation is an example of the Unicode bug.

The string "\x{61}\x{ff}" may be stored internally as either 2 or 3 octets; however it also has a length of two, consists of two characters, and will be incremented by my example code to the string "\x{62}\x{0}", regardless of the internal representation.