I think the concerns about smaller instances are valid (as I post from lemmus.org). Some additional data points to consider when evaluating an instance would be whether they’re running a recent version and the uptime of the instance.
It’d probably be a good idea to have a page that promotes these smaller instances that ‘score’ well to help distribute some of the load.
From what I understand, this is an issue with using different timestamp formats between implementations. In this case, it’s with kbin and Lemmy.