Well, as far as I remember there was a single main cap all the time, even in the first version Dave teared down.
I do think different to the fan behaviour: Unlike any PC stuff with monitoring hardware, a tacho signal from the fan is rather seldom in such hardware and even a PWM controlled (4 wire) fan is a kind of overkill to such a device. The Korad does not have to know about the speed of the fan, as long as there's a sufficent correlation between rotation speed and the supplied voltage - which is the case, normally.
And for the air flow design: If you place a fan this way directly above the heat sink, you have a powerful airflow through the cooler even in the direction of suction, as here. The cooling would be better if you change the air direction as suggested by some users, but I would never heat the components inside this unnecessary way.
Ramping speed may come from the fan regulation. Unlike the first versions with a load regulated fan speed the cooler follows the load with a load dependent delay, very normal. I remember my Voltcraft DPS2010, famous for a stupid fan regulation: the fan starts at about 42 °C and reaches may speed @ 50 °C, regulation far too steep, which results in an oscillating speed of the fan with middle loads. And, full fan speed with any current >40% of maximum. That's annoying!
As mentioned above, I have a new Welectron design model and got me another second hand, RND 320 KA3005P, answering "V5.5" on *IDN?. I have no idea how long this version is on market (and how old it is). Now I can compare both versions in behaviour and performance and do like the improvements a bit more. At a first glance, the fan speed at no load is significantly lower with the new version.
Next week I will make a "light teardown" of both to examine differences, before I go and fix the non lit OVP and OCP LEDs and the (sometimes) snarring fan.