When coasting in idle, then shifting into gear....
is it worse to shift to fourth (lower rpm) or into third (higher rpm than fourth)? Is the more abrupt change in third (e.g. 950-3000 rpm) worse for the transmission than a smaller change to fourth (e.g. 950 to 2500) with less torque?

The concern should be the percentage of speed differences between the engine rpm and the transmission rpm when the clutch is engaging them. Going from 950 to 3000 in third is more stressful than 950 to 2500 in 4th, which itself is more stressful than 950 to 1000 in 4th+n gear.
If you can't lower the input speed, then raise the engine output speed instead. Bringing the engine rpm up to close to the expected engaged rpm is another means of matching input and output speeds for limiting stress. Engine braking torque if fairly linear, unline the accelerating torque. The braking torque at 3000 isn't really much more force than the braking torque at 2500. No, braking torque isn't a concern either.
Momentum is. The intertial load of using the driveline to accelerate an engine from 950 to 2500 (or 3000) can be much higher than the braking torque depending on how quickly it is done.

Idling in gear
 

The other thing about idling in gear - even on 'standard' throttle cars (not 'throttle by wire'), there is a bypass system. For example, on my car, there is a seperate pipe coming off the air trunking before the throttle body, going to the inlet manifold.

When the load increases at idle, it just lets more air in, with the result of more fuel being injected.

Any car that didn't have this system (and ran on petrol/gas) would cut out really easily.