The idea is to recycle worldgen code as much as possible, to minimize development burden, and also the reduce computational loads on the computer, since an active site would be open, as well as what's required to spin the worldgen spinner.
Worldgen works on 1 year increments, not seasonal increments. That's why the end of winter makes the most sense. (End 1 fortress mode year, spin worldgen 1 year to sync, calculate current fortress's actions into the spin, write changes to world data, free memory, resume fortress mode.)
This might make things seem a little sluggish for world current events, but its better than the "inverted sleeping beauty" silliness we have now, where the whole world falls asleep for 100 years, except for one small outpost.
Regardless of how toady implements it, that dynamism is on the todo list, and judging from the rate of feature inclusion he's been doing, may be sooner than we realize.
Don't want to get hopes up, but I suspect toady is rather tired of fixing silly bugs with his time, and craves adding features as part of his developer diet. That would certainly explain quite a few things, like beekeepers being broken through 2 releases, while we get minecarts and other fun things added. (Especially when it has been determined what the actual underlying problem with broken beekeeping is, for quite some time now!)
I can handle a few instances of dwarves planting socks in the garden, or chickens laying bees in the nest box, if it means world dynamism in fortress mode sooner.
Again, all we have to do is be patient.