The actual example in the testsuite was more along the lines of
foo='"This is a quotes string. it'"'"'s using both types of quotes"'
-> "This is a quoted string. It's using both types of quotes"
Which is fine. That example does not aggregate quoted values with non-quoted ones.
For any potential aggregation of a quoted string "XXX" and a non-quoted string YYY, you can always write "XXXYYY" instead, until only quoted strings remain.
And I really have no idea what the Parser.HTML documentation says about it all, I do not think it actually mentions how things are parsed. But the code goes to great lengths to actually support the concatenation of strings, so it must have been intentional.
It does not follow that concatenation of quoted and non-quoted strings specifically was desired though.
And if the behaviour is not documented, it is undefined and could be changed at any time. ;-)