Are you able to explain how closed scene resolution gives you less room to improvise - eg an example of an action that you might want to declare but can't?
Sure.
First, how much organised play have to engaged in? Living Greyhawk, Living City, Xen'Drick Expeditions, D&D Encounters, Pathfinder Society? How many published adventures have you run? How many convention tournaments have you done? I've done those activities a lot, and there's precious little improvisation in those games. Or even roleplaying at times. Just a lot of chucking of dice.
3e/4e/PF really seemed designed with organised play in mind. Regimented play with little adjudication. But there's virtually zero improvisation in organised/ convention play because that goes against the intent that everyone gets the same adventure run the same way. It's meant to be your skill with the character and tactical savvy that determines your success. How the DM runs the adventure isn't meant to be a variable.
Second, is the option creep.
There are 3271 feats in 4th Edition. And 2783 general/combat/skill/ magic feats in Pathfinder.
You cannot know them all. It's impossible.
If someone wants to do some improvised action... can you be 100% sure that is not a feat? That you're not giving them a "free" feat.
Plus rules. There are lengthy codified rules on every activity. Which not only means there's less room for pure improvisation, but occasionally means you need to stop and look up the official "rule", which slows down play. Easier and faster to just use a power.
Activities like kipping up or kicking a sword into your hand or intimidating someone with a display of sword prowess. Or even using a whip to swing across a gap.
Lastly, there's how codification focuses your attention. Very often players just fixate on their character sheet when confronted by a problem. Especially in 4e with the power cards. Robust codification gives you a big toolbox, which becomes a trap for your attention. You *want* to use all your new tools, you *want* to use the options you invested a feat/power slot in, so you look for solutions using those first.
It's the sunk cost fallacy at work.