I explained it to a degree in the answer that I left here. I'm not certain if you realize how much depth your question has, and my apologies in advance for throwing a rather big wall of text as an answer to it.
It's important to clarify, we're improving question blocks substantially. There's some distinct groups of users that they affect, sorted by some interesting characteristics:
Those that tend to ask one or two questions per task they are given every day at work. Your classic help vamps.
Those that start off with several really awful questions, but would have improved had they had a shot at a few more. They can't go back and edit their previous posts because they're basically unsalvageable. These folks got off on the wrong foot, which isn't hard to do. We're quirky, what can I say?
Those that ask one or two 'meh' questions with the occasional really bad one, but only several questions per month. They end up being blocked, but it takes them a year because of the infrequency of their posts and low visibility - the system won't treat them as new users, but they essentially are. There aren't many of these, let's just call them the outliers.
The single worst offender when it comes to recidivism, or simply requesting deletion just to return with the exact same quality of questions is the first group, the ones that basically crowd source their job six days every week. Those are the users that we really want to slow down.
The second and third group aren't really hindered too much by the restriction, as they can generally start off with a decent enough question that it's lifted very quickly, or can write a few decent answers to help themselves out of it. However, those are also the groups of people that are more likely to just stop using the site due to the block than try and work around it.
So, we have one common denominator between all three groups. Due to the scale at which certain parts of the q-block system was designed, it's just too darn hard to get out of so folks either try and get around it, or just go somewhere else.
To get to your answer before I get down this rabbit hole even more, a week was chosen partly based on data (specifically, looking at medians when it comes to time lapsed between questions that didn't do very well) and a little penalty tacked on to make sure we get their attention. You could easily get the restriction lifted in a few days if you work at it, but it's no longer easier to delete your account and try again than it is to try and get yourself established and asking better questions.
Migrations are now blocked if they would put you over your quota on the target site, that just went out yesterday (and may not yet be in prod, but should be soon). Important to keep that in mind if you keep reading.
If all you wanted to know was 'why a week?' - stop here.
I'm putting something in front of the 'rougher' part of the question block system that's going to do something very similar to what the anti-recidivism limit does - it'll limit the amount of questions you can ask, based on your history, which is scoped through several different time lenses. It will start gently, perhaps limiting you to one or two a day, then a few each week, one per week, two per month and then Knuth help you because you really suck.
People that improve tend to ask fewer questions overall, and those questions tend to be much better received. Yes, we account for those times when you just have this outrageous project and need to ask more questions than usual, that's why we have several time lenses, and some other secret-sauce factoring in the works.
What does this get us?
People that treat questions as a resource that can be depleted, who learn how to ask questions only when they really need to and make them count when they do. Or, they keep throwing themselves at the wall and then get stuck in the mean hairy algorithm (Jeff wrote that, so it's intentionally both mean and hairy, I assure you - just the meanest hairiest tough-love you ever saw).
It's not yet done, but in the end, we're optimizing for giving the second group a fairer shot at showing us that they really get it and just got off on the wrong foot, and the third group a chance to either show up in the first or second group. The result should be, those that can be helped are helped, those that can't get stuck in the room with the big, mean hairy algorithm, and deleting your account no longer helps.
That, coupled with the spam layer that deals quite well with folks continuing to create new ID's trying to get around restrictions, should muffle quite a bit of noise. By the way, the spam system can also rate limit you ... oh yeah, they're gonna hate me by the time I'm done.
As of October 12th 2014, the rolling rate limits are now in place as part of the question-block system. This means, users will get more chances to write better contributions prior to being outright blocked. If you delete your account while being blocked or limited, you'll still be limited to one question per week.
While recent changes have given much greater incentive to create and build only a single account, there still must be something that makes deletion absolutely the worst option for those that hit limits or blocks.
tl;drof my answer is - it was data driven, with a bit of a penalty tacked on, and the direction we're taking q-blocks in general so they mostly only affect those that can't be helped, while helping those that can.