From a unit-by-unit standpoint, if it passes hipot today, it should pass hipot tomorrow. As long as each unit passes hipot, you shouldn't have a safety or reliability concern.
From a product line standpoint, though, you can't be guaranteed that the fan will always be built the same way. It's possible different production runs will be built differently. Maybe a different plastic will be used, or a different geometry, or who knows what. If it's not a spec listed on the data sheet, you can't assume it will never change from batch to batch.
Since asking the question, I've found out about a case where exactly this happened to us. Many batches of fans passed hipot to our 3000VAC level, well past their spec, and we shipped units with no problems. Then a new batch came in, and every one failed 3000VAC hipot. We had products due out the door, and we had to scramble to insulate or isolate our fans.
For individual units, you can be fine using unstated specs that you test on a unit-by-unit basis. But as a matter of best practice for a product line, one should not rely on unstated part specifications.
And well, roughly you could say that the amount of surface area will allow better cooling.
In combination with the amount of air flowing through/against it.
So that it will have the most possible contact with cool air.
I'm not sure what your implementation is. But I'd suggest to try passive cooling.
It really helps a lot, won't cause any noise or 'waste' energy.
(Cooling with a fan might increase the current over the mosfet and thus backfire you?)
Also, if it gets too hot for passive cooling (probably) something is wrong with your implementation.
So yes, it will work, it's effectiveness will vary on the exact implementation. But it 'should' not be neccesary to use active cooling (with a fan). The mosfet heatsinks on ebay look pretty promising and are way easier to implement and cheaper than an extra fan+temperature sensor.