Beta-Binomial conjugate prior -- pm.Binomial buggy results...?

So, there are 100 [N,L,W] sets (N trials, L losses, W wins). So pm.Binomial is receiving the N and W arrays, each of length=100. When I swapped out the Binomial likelihood for Beta-Binomial likelihood, the issue went away. What I can’t figure out is… why?

For a more extensive conversation, see my Cross Validated Stack Exchange post.