And for the fourth, something version-independent: If I hollow out some tiles next to a murky pool, will the rainwater still accumulate in there and flow over into my reservoir, or will a floodgate-less approach merely lead to rapid evaporation?
In attempt to answer some of these,
4) It can, however if its too large then evaporation will occur quickly. a floodgate probably is your best option, if not just for the extra control
In a particularly waterless environment, with occasional (and sometimes partial) pond-filling, I've sometimes attempted the following method. Controlled either by lever or pressure-plate. Although I've usually encountered other problems before seeing it proved as successful... However, on with the plan.
From the pond, set into its wall (without changing the ground 'floor' above it) a channelled drop into a single cell with an adjacent floodgate exit. Whenever the pond gets rained into, this cell gets filled to some degree or other. (You could set a floor-hatch to control when/whether it will at any given time.) And the cell is protected from the usual evaporation of water (depth-1 excepted) as it's not open to the sky.
When this cell is full (or full-enough, according to tuning) open the floodgate exit to let it fall into another pond, that you've dug, this time of more than one cell. If you have engineered it to empty a single cell of 7 water into (say) three others, then you will have at least two water in each (depending on if more than one of the original seven evaporated while the water was in transit or busy settling).
Set it up so that you can rinse and repeat this process, whenever you have enough water in any particular level's cells to dump into a large reservoir lower down. After the three-cell, gets filled to the top, or nearly so, you should be able to dump them into a set of up to nine (but perhaps better to be eight or less, due to 'wavefront evaporation') cells on the level below. A few fills from the three-cell level (perpetuated by proportionately more fills from the single-cell level, as governed by the pond's own rain-filling sequence) make this a good progressive source of water for something larger still. YGTI.
You should lose a minimal amount of the original water, per stage, until you eventually have a usable amount in a large-enough reservoir. Though it will take time, it's more evaporation-proof than trying to drain it straight into a massive reservoir, and also safe from "water crunching" floodgates, as long as you don't close them onto an overflowing lower level (in which case you've not shuffled it off to any further level).
Of course, needs preparation, and is difficult to safely (and water-conservatively) extend mid-use, which is largely why I might agree that the effort is a little disproportionate if you've got
any chance that you can get any other water system set up, or have other overwhelming problems with the fortress that are likely to cause more immediate population collapses and/or tantrum spirals.
You're also well (NPI!) advised to maintain a small reservoir below another pond dedicated to dry-season/siege-time drinking purposes, so as not to deplete the one you might be building up for whatever motives you aspire to use large quantities of water for.
I'm also not sure if any of the 'traditional' water-generation techniques that I recall being used still work. And, these days, digging a well-shaft into a flooded cavern is probably the best source of water, even given the risks and effort (and the finding of the right spot!) involved.