Ad 1. - I would really like a fix for https://github.com/stan-dev/math/issues/1495 (neg_binomial_2_log can return crazy values when phi is large) - which IMHO can negatively affect actual inferences in Stan (because that’s how I discovered it). But this hinges on getting neg_binomial_2 right and that is currently 3 separate PRs that need review and possibly other adjustments. (lbeta - https://github.com/stan-dev/math/pull/1612, binomial_coefficient_log - https://github.com/stan-dev/math/pull/1614, neg_binomial_2 itself: https://github.com/stan-dev/math/pull/1497). Not sure if it is useful to make a quick fix or wait until the other PRs make it through.
Some ideas for agenda relevant to me, not sure how much space those deserve (could be 0):
- What should be the “default” or “standard” way to check approximate equality in tests (based on discussion at `expect_near_rel` behaves weirdly for values around 1e-8)
- How much should we be worried about bugs in common-ish distributions: Do we need more unit tests for distributions?
The time is OK for me, will join.