For a statistically relevant sample of 20 rounds, an SD < 15 is pretty usable. I get as low as 10, but of course the error margin of the chrono itself is an issue at that low a number.
As to what SD leads to a degradation in accuracy, that's solvable by a ballistic calculator. 90% of your shots will be inside +/- 2 SD's of the mean , 99% will be +/- 3 SD's. Input that velocity variation and note the elevation change at a given range. That would give you a theoretical variance in group center, and therefore group size/hit probability. Depending on what you are trying to do, this would give you an idea of what's reasonably possible.
For example, my 338 LM load with a recorded SD of 10 would lead to a maximum swing over 99% of shots of +18/-23 inches at a range of 1500m. So, there is no way to be sure of a torso hit even under the most perfect of conditions, zero wind, exact range and zero hold error means the shot can still go over or under due simply to velocity variations that cannot be known until after the shot is fired.
For example
