In some ways the problem is that folks going into tech fields take the mandatory stat class. Then, when they have results to report, they pull out one of the very powerful stat software packages that are available and just plug in their numbers. They never really got a gut sense of what statistics is about. They see it as a legitimizing tool for their real work. So statistics are frequently misused, unintentionally.
So you'd classify this as a relatively recent problem?
Not only do many (if not most) scientists not understand statistics, many statiticians don't either.