As I understand the geometry, the parabola that decribes a bullet's trajectory at 100 yard is not simply a fractional segment or sub-segment of the parabola that describes its trajectory at 1000 yd. In other words, the 100 yd parabola will not perfectly overlay the first 100 yd (or any other portion) of the 1000 yd trajectory. That would seem to be at least part of the explanation why loads developed at 100 yd might not be optimal at some other distance.
One observation I would make regarding the 1000 yd/100 yd group data presented above is that the groups at 100 yd are not even close to what I would consider "good" groups for a well-tuned F-Open rifle, ranging from about 0.8 to over 1.0 MOA. The possible exception to that statement would be the 100 yd target from the final pair (B41), but even that 100 yd group is still larger than 0.5 MOA. Thus, I think trying to infer anything from the shape of the 100 yd groups that could be used for comparison to the 1000 yd group shape would be difficult. Because of the relatively large 100 yd group size, I suspect there is too much uncertainty in 100 yd group shape/shot placement to reliably compare them to the 1000 yd group shapes. It may even be possible that the resolution of the e-target system may come into play with the 100 yd groups. In other words, there are potentially a number of sources of errort that make interpreting the shape of the 100 group more challenging. Having said that, the final pair of groups at 1000/100 (B41) are both similar in shape, both being generally roundish. Further, had it been centered on the target face, the 1000 yd B41 group would have been fully contained within the 10-ring, with very close to a 50% X-count. From his statements, Tod has obviously seen a lot more of these targets than he posted above, which can make interpretation of any trend much easier, so I fully accept his assessment of the results. Nonetheless, I think it would be interesting to see how a load that shot much tighter at 100 yd (i.e. 0.25 MOA or less) might behave at 1000 yd.
In any event, one of the issues we have to deal with when assessing groups at 100 yd is one of resolution and/or measurement. We can add to that that the bullet hole diameter represents a much larger proportion of the group spread than it does at 1000 yd. Increasing the distance is an easy way of "stretching out" the Y-axis, thereby effectively improving resolution. However, that also increases the potential effect of the ambient weather conditions on bullet trajectory. The major benefits of carrying out preliminary load development at 100 yd are minimizing any conditional effects, and that most shooters have ready access to a 100 yd range. Not everyone can do load development at 1000 yd. So I will end by circling back to my inital statement regarding the different shapes of the bullet trajectories (parabolas) at 100 and 1000 yd. If one could do the kind of analysis Tod has shown above and reliably detect a certain group shape at 100 yd that was associated with a load known to be tuned really well at 1000 yd, then it should still be possible to get the load pretty well developed at 100 yd. Even if the 100 yd group wasn't good/tight/tiny, as long as it had a recognizable and reproducible appearance, it could be used to tune a load that would shoot well at 1000 yd. I don't think anyone would argue against the benefits of doing load development at the distance to be used in matches, but not everyone has access to a 1000 yd range for load development, and/or the e-target setup to facilitate that process. The more one can do at the distance/range they do have access to, the better off they will be. This also raises the question that if groups fired at 100 yd offer little to no insight into optimizing groups at 1000 yd, is there some other intermediate distance, such as 200 yd, or 300 yd for example, commonly available to most shooters, where the results do better match those obtained from the same load at 1000 yd?