Moonsinger's simple strategy would be better than it stands at present (in my opinion), but I don't think it would have the kind of impact on reduction of micro-management that I would like to see... I would be in favour of almost any system that alowed shields to carry over.
But I think Rubruk has hit on a particularly sound idea, in that it should be allowed if the units are queued. I also believe that being able to build four warriors in one turn with a 40-shield city is perfectly acceptable.
In Civ2, I recall there was a 50% penalty when changing between the three main types, but only on harder difficulty levels. I personally am glad this is gone, although I admit it might be more realistic (and more challenging) to keep it... All I would say is that I emphatically argue to keep it (mostly through laziness, so that I can decide a few turns later what to actually build), and my reasoning would be that many improvements use the same base resources. In the earlier eras in particular, imagine the 500 years it takes to build a barracks (assuming the very beginning, so 50-year turns, collecting 4 shields per turn). The first 450-odd years of that would probably be collecting raw materials, such as wood and stone. The final 50 years (the last turn) would be the period where those materials are turned into parts of the barracks. If, after 300 years, it was decided that a temple should be built instead, then much of the raw materials collected would still be of use, they would just be made into different things. This probably holds true for converting to wonders as well, although I admit things become a bit harder to argue when converting a barracks to an archer...
I also agree that science should be carried over. I thought this wasn't the case back in Civ2, so I always used to do the science-slider technique back then too! What a waste!