Competition and misconduct! This is basically what the current developments in science politics boil down to. This convergence in opinion comes from multiple sources, such as Coturnix, 3 quarks daily, Janet and an interview with Gerhard Fröhlich in the German periodical Laborjournal.
What is this all about?
In this current era, there is an overproduction of scientists and a widespread shortening of positions. Similar to other economic areas, investments are being made for technologies rather than labor, leading to a shortening of research positions with stable or increasing budgets. This trend is aggravated by funneling research funds into a few, highly competitive areas.
What does all this mean for 299 out of the typical 300 applicants for a faculty position: you're now between 35 and 45, you probably have a family to feed, you're highly trained and extremely specialized (sometimes all you know is your molecule and the ones it interacts with), you've never worked in a 9-5 job outside of academia. Now doesn't that make you a fantastic candidate to compete with fresh graduates on the industry job market!
But of course, you can save yourself and your family the choice between poverty or dual jobs at McDonalds and as Taxi-driver: all you need to do is get a Nature paper or two and maybe a Science paper, too, just to be sure. This will surely beat the other 299 candidates! Unfortunately, science cannot be planned like this, so potentially brilliant candidates may just have had bad luck and never get a high impact publication. Or maybe if only this one experiment would come out a certain way... Obviously, the temptation to cheat is growing with the overproduction of scientists and increasing competition and we will witness an increasing flood of unreliable studies as the competition increases.
Increase the incentives and cheating will follow. This is the rule for sports and will be the same for science. With the economic impact of science, can we afford to waste tax-payer money first on training too many scientists, then by paying several times for the same research by rewarding few, fashionable research topics and then a third time when resources are wasted trying to replicate flawed results?

What to do?
"Open science" is one way of alleviating the effects of increasing competition. But several other measures have to be implemented as well. Effective punishment of course can be implemented as deterrent. One will have to experiment as to the effectiveness of such measures. But in my opinion, the incentives to cheat need to be lowered as well. Fewer scientists should be trained and the selection has to be implemented at a much earlier career stage. Alternatively, more positions should be created, but that option is more difficult to implement than working on the supply end.
Maybe one should run a few simulations using game theory to find out where in parameter space "all defect" ceases to be a viable ESS?