As I look back on my career, there have been things that I learned early on that I still am able to leverage today (i.e. bash, linux, vim, java). There are a slew of other technologies that didn't last (i.e. DCOM, Corba, J++, J2EE). If I'm going to invest my time in learning, I'd like to realize some return on time. Granted there are cool experimental technologies that I learn with the realization that they may not be around in a couple years, but I can plan/architect around that.
Those specific technologies may not have lasted in those formats but OO techniques and patterns in them will persist for a long time like ModelViewController and its variants.
There may be some cool new tech that comes out with a neat pattern for solving a problem, we need to identify that pattern and re-use it.