Update on my earlier post. It seems that there are bugs running aspell on UTF-8 code via ispell.el.

For example, with the configuration I listed in my first post, running M-x ispell on a buffer containing

Grüß Gott       <- misspelled English with UTF-8 chars
“aBx”           <- misspelled English inside UTF-8 quotes

produced an error

Ispell misalignment: word `BaBx' point 58; probably incompatible versions

when aspell caught on “aBx”.

This seems to have been a problem for a while (2003 bug report and a possible 2005 patch against the 2003 ispell.el-3.7beta), but I don't know enough about emacs to want to go that deep. For now I will just run aspell -v <my-file> from another terminal and revisit the file from emacs afterwards, or I will highlight non-UTF-8 areas and ispell them seperately.