I think 45 in the dressage is reasonable. I think 40 might not be. An example: My only score worse than a 45 came in a monsoon, when the footing deteriorated and we were slipping through corners (in hindsight, I should have put in studs for dressage). That score would not have counted (and I ended up withdrawing given the conditions). I already tend to be tense in the dressage, with my scores on that horse mostly in the 34-39 range, with the occasional 40-42. That horse was a total genius in the XC, though, easy to gallop and easy to make time. Should some of those shows, which were MERs, not count? Consistent scores over 45 should be a red flag, though.
I think every horse is allowed the occasional stop–a slip, a deer running in front of a fence or a jump judge opening an umbrella, a rider deciding discretion is the better part of valour and taking a circle. I wouldn’t say a horse needs a perfect record, or even a percentage; a green horse might have two shows with baby moments, then four perfect runs, by the end of which he is more than ready but still only has a 2/3 clear rate. Or a combination that moves up, then moves back down and then back up–do the old events count against them? I do think that a RECENT record of shows with stops–more than one on a course or more than one show with a stop–should be a red flag.
I agree about the time; an MER should not have more than a few time faults (and frankly, it shouldn’t really have any, but of course there are always exceptions–bad weather, problems with timing after a hold, etc., so I also have no problems with some flexibility here). But consistent and significant time faults should be a red flag.
I agree about the SJ; consistently pulling lots of rails should be a red flag. Then again, if you get into the triple poorly and pull a rail at each fence, you have three down and only one bad ride. So here again we should look at the bigger picture; did the rider have consistently clean or mostly clean rounds, or did the rider have 3-4 down at every show? There is a difference, and the latter should be a red flag.
And even if we analysed each rider’s record to account for all of these things, they would still only give us part of the picture. It’s better, but it’s still not a qualitative assessment.