I actually prefer a larger range with a lower multiple, and here is why. Presuming that each digit on a 20-sider represents a 1 in 20 chance of hitting (a slight simplification due to variable bonuses, I believe), taking a weapon with a 2 digit threat range nets you a 1 in 10 chance of critting each time you swing (presuming you hit with a 19, which you hopefully will). Each swing with a 1-digit threat range nets you a 1 in 20 chance of a crit. This means that on average, it will take you 20 swings to get a crit with an axe or scythe, and 10 to get one with a longsword.
From this respect, scythe and longsword would look equal presuming they did the same damage (scythe does 2-8 rather than 1-8 providing a slight advantage), while an axe would fall behind. As shown:
Longsword: [1-8 damage (average 4.5)*crit multiplier (2)]*chance of crit (.1) = .9 average crit damage per swing
Scythe [2--8 damage (average 5)*crit multiplier (4)]* chance of crit (.05) = 1 average crit damage per swing
Axe [1-8 damage (average 4.5)*crit multiplier (3)]*chance of crit (.05) = .675 average crit damage per swing
The scythe is sort of an apples and orange comparison due to the two-handed factor. For further comparison, here is a heavy pick and a rapier:
Heavy pick [1-6 damage (average 3.5)*crit multiplier (4)]*chance of crit (.05) = .7 average crit damage per swing
Rapier {1-6 damage (average 3.5)*crit multiplier (2)}*chance to crit (.15) = 1.1 average crit damage per swing
As you can see, for pure damage dealing reasons, crit range is generally more important than multiplier. A second reason is overkill. Each point of damage you do beyond what your opponent has is wasted. You are much more likely to waste your damage on a dead opponent if you are dealing out a x4 crit multiplier hit half as often versus dealing out a x2 crit twice as often.
It's also interesting to think about the effects of putting "keen" on a weapon versus getting a normal +1 on it, looking at the equations above.
