Gregory's Server
@SK53 Yes I can imagine that would be a problem for scaling up an analysis like this. My sample here is small enough, and I know the area well enough, that I was able to look at the changes individually and see that none of them were due to remote mapping. There were some manual typo fixes but that's inevitable, there is no way to check the name of a shop especially if it's not a brand (and most of these shops are not).
@SK53 I think I have more or less reconciled myself to check_date, Everydoor also uses it.
I had my doubts originally as it feels a little odd to have this kind of metadata in the tags. But I have found it has significant practical benefits when trying to systematically survey an area. I now realise that in the past I had been repeatedly visually confirming some shops while ignoring others. Even the side of the street I tend to walk down turned out to be more biased than I thought!