The evidence of progressive correctness comes from the various large areas I have worked, for example, the Frankfurt and Prague and Vienna Jewish communities. I know it is hard for people to believe or understand, but there are many people (not just me) who are extremely active on Geni, doing more than you could ever think is possible. Some of them become curators (there are several hundred volunteer curators now) and we have frequent discussions (some public, some private) on how to resolve problem areas, for example, when recurring mistakes creep in. One of the tools we have as curators is to lock down profiles so that new relationships cannot be added without curator approval. The tree is always a work-in-progress, but it does improve over time. Look, for example, at the tree of the Maharal of Prague. There are still some issues, but we have tried to eliminate much of the apocryphal ancestry that you will find on most other sites, based on recent scholarship. The more evidence is added to the profiles, the easier it is to avoid and resolve errors. Slowly, but surely, the tree comes together in ways that just are not possible when people work by themselves. In the Frankfurt tree, I collaborate with a number of others who scour the tree looking for conflicts and discrepancies. The Geni program also has features that can alert you to potential errors. In the Prague Jewish tree, I have fewer collaborators, but we are gradually piecing together the families, using old and new resources, so that when people find it, they do not have to reinvent the wheel and can take advantage of the work that has already been done., and then make improvements. We all make mistakes, of course; that is unavoidable. And anyone who has experience working with old records knows that there are often errors and ambiguities. As I have written previously in one of my old blogs, there is no such thing as certainty in genealogy. We are all just doing our best to figure things out until the next piece of evidence comes along that either confirms our assumptions or requires us to reevaluate them.
I agree that it is one of the things that is very difficult for users operating in a smaller field to understand, that a process that allows some errors to creep into their tree is also a process that ends up catching and correcting even more errors. I try to explain that it is like a machine that goes three steps forward and one step back, three steps forward and one step back . . . . Over time, it averages out to real progress.