There are essentially two ways to set the spoke, or deform metal generally:

Bending past the yield point. This is what I do building wheels, because it is simple and quiet. If the spoke needs 20 degrees of bend, you bend it maybe 50 degrees and it deforms to about the 20 you wanted. The reason for the way this happens is that the elastic deformation never stops happening, even after the plastic yield point is reached. So the spoke has to be flexed enough to spring AND bend the spoke at the same time. This method is not exact, though you can make each of the bends equal to each other. The precise amount of bend is less important than just getting the elbow to an angle where it seats against the flange before there is any real tension.

Bending can be done during lacing or after, since untensioned wheels have a fair amount of slop in the spoke length to push laterally.


The other technique is impact. This is the method Sheldon Brown promotes. It is arguable somewhat precise, as it uses shock to 'cold forge' the bend to the shape of the flange on an already tensioned spoke. This is the technique I use on a wheel I suspect was never seated, and I am not going to zero the tension out and come back up. Aside from the noise and extra tools required (I use a mallet), I don't like to do it to new wheels out of respect to the customer's hub flanges, since smacking the spoke uses the aluminum hub flange as an anvil, and aluminum is not good anvil material. Does it matter? I doubt anyone has ever been able to study what long term effects it has on later flange failure - so do whatever you want. Mine is just a preference.


In my experience, a quality machine built wheel that has broken a few spokes at the bend will stop breaking spokes if the existing ones are set with a mallet and then the tensions are evened out and put in the proper range for the kind of spoke and the number of spokes on the wheel. (Higher spoke count wheels should use less tension per spoke as the total wheel tension matters, and higher tension puts more localized strain on the rim's nipple holes. No reason to strain a well supported, high spoke count rim for no reason.)

It may be that simply correcting the sprung bends is enough to buy a lifetime of fatigue resistance, which might be born out by the observation that even chain damaged spoke bends seem to survive ordinary use. If bends, regardless of being sprung, were major stress risers, damaged spokes would break all the time. And they don't.

Or, it may be that striking the spokes with a mallet not just deforms them, but stress relieves them, resetting the grain and removing a portion of the prior fatigue. Someone with a lab could do that sort of analysis, but it really isn't important. I also blacksmith knives, and it is often said that striking the anvil next to the work object stress relieves it, but I have never bothered given all the heat cycling.


So anyway, if you are breaking elbows, get someone to smack those bends and re-tension with stress relief. In my long professional experience, the spoke breakages will stop right there and no new spokes will be necessary.


On tension, the most important thing is uniformity rather than a particular tension. So use a tensiometer to get in range, then fine tune tension by tone, like a piano tuner. Your ear is more accurate than the collection of pivots and springs in a mechanical tensiometer. Perfect uniformity is usually not possible while achieving a straight rim, but uniformity allows you to see how to shape the rim with multiple spokes having influence over the same rim section. You can have a very straight rim with very uneven tension - especially stiffer modern rims.