DrDiettrch - totally agree with you on the gate capacity issue. Most of my power FET work over the years has been switch mode power supply design / development. The trick here is to pick a FET with a high enough current rating to do the job, but not too much over, as per your comments, the gate capacity being higher with larger FETS. Most of my time in selecting FETs for a particular task was doing the trade off between current rating, RDS on figure, and gate capacity / gate charge figure. Miller effect can also have big effect on rise-time. (Drain to gate capacity). The numbers can all change with voltage rating - the higher this is, the harder to get a good compromise on RDS on and gate charge / capacity, so voltage rating should not be higher than needed.
Once a suitable device is chosen, hit the gate as hard as you can to turn it on quickly, pass through the linear region as fast as possible and fully saturate when on. The RDS figure can vary significantly over the range of drive levels.
To turn on fully and fast you need current as well as voltage. A good driver for switch mode use would typically have 2 - 4 amps available.
Having said all that FETs have come a long way over the last 5 years or so. Very high current FETs (50amp class) now have gate capacities and charge figures that are very low compared to early offerings. Typical (non-isolated) switch mode efficiencies can be around 98% - 20 years ago when I first started working with these things around 90% was considered good.
Whatever is good for switch mode operation should also be good for PWM, but probably not quite as critical as frequencies are generally lower. At very high frequencies rise and fall times can become a significant part of the total cycle time, lowering efficiency.
Any how that's my 2 cents worth.