@icing I think they were doing some funky stuff where they restarted parts of their stack but tried to avoid full reconnects/downtime? We're now back ag no drops, no out-of-order, just regular train amounts of jitter

(but I'm here for four hours and bored, soooo)