In this discussion, I see the first warning, maybe it is related to the installation.
Have you tried to fit the model with a simulated dataset? My experience when I try a new model is to fit it with a simulated data (with varying sample sizes). If Stan works properly, it means that the reason is not because of Stan.
P/S: If I am correct, specification by, e.g., y\sim dist() is equivalent to using target += dist_lpdf. However, for me, the former is more readable.