Tenure track faculty receive no training as teachers, and whereas they engage in intensive, daily, professional development activities with regard to their research, they typically receive only the lightest mentoring concerning their teaching, and they receive it from people who, themselves, have had no training and for whom neither teaching nor mentoring new teachers have been part of their professional development.
On this matter I have become something of a zealot, the end process of having a difficult time arranging a classroom in a way natural to me and satisfactory to students, and having talked many (,many) people through the same process. A (now successful) colleague in grad school once remarked that the academic job market is a strange one. One is hired to do a particular job--teach--but one's hiring and promotion depend almost entirely on research. While that is certainly stylized and not entirely true, depending as it does on the type of higher education institution one ends up at, it does get to one of the central oddities of the grad school process, namely the distinct emphasis given to research over teaching. I learned after a few years on the job market that expectations will be both high and low: most jobs will assume that someone who possesses a PhD will be able to generate decent new syllabi for any course reasonably within their field of study given no more time than a summer, but virtually no experience other than serving as Instructor of Record will count as 'actually' teaching.
While the assumption of general competence is usually correct--it doesn't take that long to master new subject matter once one knows how to research--the idea that teaching is just a natural outcropping of the research process is exceedingly strange. To see this, one need look no further than people who are grading for the first time, who are usually plunged into a state of existential terror by the prospect of being objectively wrong in their assessment of a student's work and of properly calibrating the grading scale for the entire class. It applies to matters great and small: putting together a lecture (hearing hundreds or thousands of lectures is not much of a help unless you already know what to be listening for), writing assignments, determining the right balance of out-of-class work in a syllabus, running a discussion section (especially in a class where controversial subject matter will come up), and on through such basic questions as where does my authority in the classroom come from? and into such intricacies as handling plagiarism, mental health issues, and failing students. The average instruction given to the average grad student about this is minimal. Though I believe I was (and remain) a good instructor in the classroom, I got there by cobbling together the information my own professors would periodically share about their teaching, through the luck of TAing for the same courses multiple times (and so being able to think about what made the professor so effective and beloved rather than making sure I followed the content), a healthy amount of outside reading on pedagogy, and notably crashing and burning my first time as Instructor of Record. (It was not a failure, for the record. I did fine. But I made a few missteps, did not enjoy myself by the end, and it showed. I would also completely revamp that syllabus.)
The biggest leap came from finally getting access to the teaching and learning opportunities available to tenure-track faculty. By that time I had the combination of experience and theory necessary to take control of the courses I was teaching, and the general teaching-and-learning material that came into my hands could be put to use; most crucially, I knew when to put aside the advice of others as being unfit for what I was attempting to do. The solution to this problem is more intensive focus on teaching at the earliest stages (where there is more time and the incentives are different than once one is on the job market), hand-holding through the first attempts and gradually increasing the skill set of grad students as time goes on. Duke had a few of these programs when I was in my PhD program, and I did not take advantage of nearly as many as I should have--but more are needed.