I am 100% certain that Matthew is correct. Once I have enough espionage to know what techs the AI is currently researching, I often abandon my current project to research a tech the AI is just beginning. That way I can trade it to them a couple of turns before they actually research it and loose nothing of the accumulated research on my original project.
You keep all you researched. It's totally possible for instance to switch research project when your survey ship nears an anomaly, get a bonus in that tech, and switch back to the previous one. You'll get the bonus from the anomaly for the tech you switched to, and get your end of turn research in the other.
I used to but no longer, mostly because of the tech anomalies, but also the space sharks. In fact, I tend to use auto-survey to check where the nearest anomaly is, and then switch it off.
On the last turn researching a tech, what happens to the left over research?. Is it thrown away, kept for the next tech, or given back to you in the form of cash?.
For example, say you produce 60pu in science each month and there you start researching a tech that costs 320pu. After 5 turns you'll only need 20pu more to finsish the tech and you hit the turn button. What happens to extra 40pu?. Do they get thrown away?. In that case isn't the game forcing us to manually change the reasearch spending on the last month from 60pu to 20pu if we want no waste...?
If the research tech is wasted it's like paying your scienctist to be on vacation. This makes no sense. A research lab manned by a fat security guard has to cost less to mantain than a research lab full of busy scientists. Same goes for planets that produce nothing in empty factories, they should spend no money! (or at least a small fraction of what it would spend if it were busy).
Leftover research is wasted. When you are tight for cash, it is a good idea to try and set the slider for optimum research. Later on in the game (when you are filty rich ), you can just set a percentage and forget about the waste.