The ability to override the character set was removed 2 years ago, in bug 1713786 (and perhaps others). Several users complained already at that time about mis-detection being imperfect / insufficient (including myself). An implicit justification to the removal is that mis-detection is either already mostly-perfect or would be perfected quickly enough to not merit supporting character set forcing.

Two years have passed, and this has not happened. There are numerous situations in which the character set misdetection fails (even if they are much less frequent today).

Character set forcing should therefore be restored. And even if I were to provide one, or two, or five examples - there are still likely more than I know about, with other languages and other encodings; it is - to be honest - hubris to believe that such a heuristic can be relied on perfectly.

... and if even this does not convince: There are some situations in which it cannot succeed even in principle. Those are messages which contain text in more than a single character set.