You mean IK constraint stretch, not mesh scaling. It's true IK softness > 0 disables IK stretching. I agree it would be better to support both at the same time. The issue is the math is difficult, depending on what behavior is acceptable.
Softness slows down IK when the bones are near straight by essentially pretending the target is closer. That means you don't reach the target even when it would be in range without softness.
Stretch scales the bones to reach the target. We can only stretch once the bones are straight (otherwise it gets extremely complex).
If we allowed stretch and softness at the same time, we still can't stretch until the bones are straight. That doesn't happen until the target is some distance beyond the reach of the bones (dependent on softness amount). Then when we stretch there would be a jump where it suddenly scales to reach the target. We could instead say softness makes it not reach the target, then there would be no jump. Or we could do some catch up scaling and reach the target only when it's even farther away, but this seems complex and the faster scaling to start may be unwanted.
I tried it (not reaching the target) and it works OK. Softness slows down the IK bone rotations as the bones straighten, so there's a slight pause before the bones are straight and stretch scaling kicks in, depending on how much softness you are using. It can't fit into 4.3 (sorry!) but we'll fit it into the next version of Spine.