Ampyx Power: Wind energy systems using tethered wings above 200 meters

A windmill that isn't rotating yet is an entirely different device than one that is already turning. Once moving, indeed, the three bladed device with a gearing has the same torque as one that has many more blades. But when it isn't turning yet the device with more blades has far more torque than the one with only three blades and that is why it can get started at all.

Pumps are particularly unforgiving, most of the designs out there will have an uneven load distribution on a single rotation of the outgoing shaft to the pump and so will get stopped at the point where the required torque is at its maximum. Precisely because of the distinction made by the questioner: why does so much of the surface allow air to 'slip through'. Which it will in fact do when the mill is standing still or still moving very slowly.

So on a three bladed device the start-up torque will be a small fraction of what it is on a device with many blades, which prevents the three bladed device to start rotating effectively stalling the rotor until the amount of wind is high enough to overcome the initial resistance of the mechanism, which can be quite high. In low wind situations this will allow the multi-blade device to produce far more yield, in moderate wind situations it will produce somewhat less yield and in very high wind situations it may cause the three-blader to overspeed and destroy itself unless it has a very good furling mechanism (such as variable pitch). On a three blader the only parts active to generate this start-up torque are the three blade roots out to about 1/3rd of the rotor diameter.

This limitation applies to purely mechanical direct drives without any trickery such as start-up relays or clutches that only engage when the mill is spinning at a sufficiently high rate, in which case the rotor can stay unloaded long enough for it to get into a high enough rate of rotation that a torque convertor such as an electric motor or gearbox can be engaged. But that is a far more complex setup, one that waterpumpers that run for decades without maintenance can do without.

For electricity generating windmills the situation is a bit different in that most of these have an output voltage below which the energy the mill produces isn't used at all, no current will flow until the output voltage exceeds this lower limit. This effectively acts as a clutch that only engages when the mill is already in an efficient domain of operation. MPP tracking can help extract some power below that speed, something that may be useful in domains where there are long periods of low speed wind.

Just because there wasn't a good computational model for a long time doesn't mean there wasn't a wealth of practical knowledge about windmills empirically gained over three centuries. We already knew what worked and what didn't, we just didn't know why it worked, and we weren't able to scale up the smaller machines to the sizes we see today. The latter was also heavily affected by the development of stronger materials such as carbon fiber.

At a guess you have a highly developed theoretical knowledge about windmills, but not a very well developed practical knowledge, as in actually building some of these models to verify your theoretical knowledge. If you had you would not make such a simple error because you'd know from actual observation that this difference exists in practice and why it exists, rather than just to make a claim from what you know from your - in this case faulty - theoretical knowledge.