This is an unusual grid. As you can see, it has a set of quad-stacked 15s running through the center of the grid, and a pair of triple-stacked 15s at the top and the bottom. When I first started constructing this crossword, I did not have this goal in mind. This is because all of my quadstack puzzles have been constructed "backwards." By "backwards" I mean that I start by trying to create a set of four 15-letter words to stack independently of the grid.
But, how are the actual quad-sack sets constructed? The answer is: by any means possible!
Some solvers think it's 100% computer and word list management. However, in reality, using a computer to "brute force" a quad-stack is often doomed to failure. It is possible, but the results are usually very poor, because the computer word list has to be very, very large. And that means: lots of unusual, really poor, rare words. Furthermore it's a good way to fry your hard-drive! (I speak from experience!)
The best approach I've found, is a combination of old-fashioned hand constructing, with varying degrees of computer assistance.
When I think I have what may be a viable stack, usually with at least a couple of nonsensical 4-letter combinations crossing the stacks, I set to work trying to design a possible grid (if I'm lucky) around the stacks. Frequently I do this on my iPhone. Indeed, some of today's puzzle was actually completed in the parking lot of a local late-night donut shop ... and heck, I got free police protection too ;)
Seriously though, as I said, these puzzle are a combination of computer assistance, hand construction and sheer luck. And in the case of this puzzle, it was sheer luck that gave me the possibility of adding the extra stacks.
However, if I had not found two strong 10-letter words (CHILD BIRTH and TORE TO BITS) to join the stacks, Will Shortz would not have accepted this puzzle. Those two long words meant that Will was willing to break his moratorium on this style of grid, mainly because there was greater interconnectivity between the stacked parts of the grid.
Personally, I still think the puzzle (obviously) would be improved by a less segmented grid. But, to the best of my knowledge, this is the first time a 3/4/3 stack pattern has ever been published. So if some solvers think that this is too much of a "stunt" grid: I plead guilty. But I do hope this is still interesting and entertaining to solvers.