Why do BLDC motor (1 kW) controllers have so many MOSFETs?
The reason to use multiple MOSFETs is to lower power dissipation resulting in a cheaper design.
Yes one MOSFET can handle the current but it will dissipate some power as it does have some resistance, typically 9 mohm for the IRFB3607.
At 25 A that means 25 A * 9 m ohm = 225 mV drop
At 25 A that means 25 A * 225 mV = 5.625 W of power dissipation
A heatsink for that would need to be substantial.
Now let's do the same calculation for 4 IRFB3607 in parallel:
Now 9 mohm is divided by 4 because of 4 parallel devices:
9 m ohm / 4 = 2.25 mohm
At 25 A that means 25 A * 2.25 m ohm = 56.25 mV drop
At 25 A that means 25 A * 56.25 mV = 1.41 W of power dissipation
That 1.41 W is for all MOSFETs together so less than 0.4 W per MOSFET which they can handle easily without any extra cooling.
Above calculation does not take into account that the 9 mohm Rdson will increase when the MOSFETs heat up. That makes the single MOSFET solution even more problematic as an even larger heatsink is required. The 4 MOSFET solution might "just manage" as it still has some margin (the 0.4 W could increase to 1 W and that would still be OK).
If 3 MOSFETs are cheaper than one heatsink (for dissipating 6 Watt) then the 4 MOSFET solution is cheaper.
Also production costs might be slightly lower for placing 4 MOSFETS compared to 1 MOSFET + Heatsink as the MOSFET has to be screwed or clamped to the heatsink, that's manual work so adds cost.
An added benefit is that reliability becomes better as those 4 MOSFETs are by far not "worked" as hard as a the single MOSFET.
Could we use a "4x" bigger, 2.25 mohm MOSFET?
Sure, if you can find it ! 9 mohm is quite low already. It gets increasingly difficult (and more expensive) to get lower as the influence of bonding wires comes into play. Also for sure four "middle of the road" MOSFETs are cheaper than one big fat MOSFET.
For almost all electrical components, lifetime decreases exponentially with increasing temperature. This is especially true with capacitors, which are found in BLDC motor drivers to decrease electrical noise and high-current peaks.
Let's say that the controller with 4 FETs per phase increased in temperature by 10°C at the rated load. Assuming an ambient temperature of 30°C, the controller would be running at 40°C. At this temperature, even standard-temperature range aluminum electrolytic capacitors could last over 120,000 hours.
If the same controller were to be built with 1 FET per phase instead of 4, the resistance would increase by a factor of 4 and the I^2R losses would also increase by the same amount. With the same heat-sink, the controller would experience 4 times the heating above ambient. It would now be running at 70°C. This would cut the lifetime of the capacitors by around a factor of 10, and would also decrease the life of other components similarly. To counteract this, a larger heatsink would be required, and it would be cheaper (and smaller) to just use more FETs.