Chains vs iterations behavior in Turing sampler (conditional priors)

No problem. The example above now works. Unfortunately, the error was quite stupid: I intended to pass data to the model, but I passed an old variable y, which was lurking in my session. I was discussing parameter recovery with someone on slack and they said that parameters are often difficult to recover in mixture models.

Evidently, the Turing website has changed. You can find some more info here:

You can also find information about identifiability here:

https://mc-stan.org/users/documentation/case-studies/identifying_mixture_models.html