Well, LaTeX (or TeX for that matter) can do all kind of ugly things if asked gard enough to do so. What you describe (without giving an example which makes it a bit difficult to do more than guessing) since to indicate that the class file for that journal is badly broken probably due to a misunderstanding or two by the person who wrote it.
Here is my guess on what happened:
TeX has two parameters \lefthyphenmin and \righthyphenmin that define the minum length of characters that need to be at either end of a word after hyphenation. The default values for these are 3 and 2 so words with just four characters aren't ever hyphenated if the default is in place.
Now you can increase either parameter, e.g., by asking for \righthyphenmin=3, however you can't actually decrease the value without producing chaos — and that is what happens in your case (the \righthyphenmin must be set to 1 somewhow else you couldn't get a trailing single character on the next line).
The reason why you can't decrease from the default is simple: all the standard patterns for most (or even all?) languages have been build under the assumption that the minimum values are 3,2. Thus, if you set it to 1 then the pattern may allow breaks with just 1 character simply because they have no rules for preventing it because that is assumed to be an impossible case (makes the pattern much smaller).
In theory one can build pattern that work correctly with 1,1 (and if a language has a few words that actually allow a hyphen break resulting in a word part of one character they would work correctly, but the pattern set would be very large to prevent all the bad ones, which is why this isn't done.
So my guess is that the class actually set the min values to 1 under the mistaken assumption that this helps linebreaking in small measures without doing any harm. Well it does help but it does a lot of harm :-)
So if my assumption is correct, then the class should be corrected.
;-)(I know it might be confidential). But can mask out everything except this and try. Or try to reproduce such an issue from your local pc and post it here. IMO, that would attract more readers.microtype. I had a similar issue in my paper, and it improved it.