As a reviewer noted nicely:
User is editing too much post today. Obviously not interested in content but in reputation
Not exactly a valid rejection reason in general, but said user suggested many edits indeed, and while a few were ok, many merely placed keywords in backticks (which btw should not be done for non-code, e.g. for the reasons listed here) and should have been rejected (a shockingly large amount of these were however approved). This suggestion behaviour is most certainly not what the system is intended for, so I wonder if there is a limit on suggestions in place? Note I'm not referring to the edit-ban due to rejections for seven days but to a general limit of, say, 20 suggestions per day.
As many of you have noted, reviewers actually accepting such crappy suggestions are a problem. But they are not the problem, but rather a consequence. Limiting the amount of edit suggestions1 limits the source of
all evil bad edits, while the serious reviewers are rather the victims who loose the battle against are slower than robo-reviewers and thus more likely to reject a suggestion when it has already been approved. We should educate the editors to "make it count", which to my knowledge is already the desired behaviour, but not in a vague "minimum character count" - one missing backtick edited in may severely improve a post, while a second one to emphasize a non-code keyword is bad.
1 And as discussed in comments, 20/day is not necessarily a good measure, something like max. 5 unapproved edits in the last 24 hours where approval a) make another edit available and b) increases that limit with time might be better - the suggestion here is about the consideration of such a limit, not about its implementation details
On an unrelated note, please don't [approve suggested edits](link_to_suggestion_here) using backticks for emphasis, but reject or improve them - see e.g. [here](http://meta.gaming.stackexchange.com/q/7437/88) whycomment on one of their posts (unfortunately the only way to call out to them). It's not exactly model behaviour by my side I assume, but fortunately users are understanding