In 'unicode_520', for example, the two characters "ae" are treated as equal to the single character "æ". Ditto for the 'german2' collation. This is an example of where the the 'general' collation is a little faster -- because it takes less effort.
In most collations "lj" = "lj", but not in 'general'.
A longer list of such differences can be found here:
http://mysql.rjweb.org/utf8_collations.html .
There is a third category of collation, 'bin', as in 'utf8_bin' and 'utf8mb4_bin'. These are the fastest, since they blindly compare bits; no case folding; no accent stripping. That is, "A" != "a" != "á", etc.; in all(?) other collations, those are treated as equal. (MySQL 8.0 promises to have some collations that will invalidate this last statement.)