I like the idea of each site researching it's own thin.g It doesn't look like that is the direction they are going, but that is a very cool and fresh idea. I'd love to see it in the game.
I think this is a good idea and it has absolutely nothing to do with losing buildigns or losing research.
Here's an example implementation:
You have research sites 1, 2 and 3 with an output of 10 research each.
You could have 30 rp focused on a single tech (single tech research model).
You could divide these points however you like, but that would always be less effective than putting all in one place in sequence.
Or you could assign each site to one research, and each tech would only get benefits from the best research site. This means to double your tech output you must research two techs.
Now the downside is you can't concentrate on a single tech and get benefits. That can be circumvented by using randomness or diminishing returns:
Jsut change the output so it's random: Sites give 5-15 output.
If you put all sites on researching the same thing, instead of researching one tech with an average of 10 points per turn, you receive that tech with an average of 12* points per turn (* ok it's not 12, it depends on the random number distribution and I'm not going to do the math, as results would be very different if the randomness is a bell curve or linear...).
There are other ways of increasing the average without using randomness, but I think random is cool as it illustrates the serendipity of technological advancement. For instance, the second best site only provides 1/2 its output, the third 1/3rd, etc.