An issue with the shape of a continous latent that depends on a enumerated discrete inside scan

It looks like this model is not supported. There is a relevant discussion here Modelling mutivariate switching dynamics - advice on getting started?