The current implementation means that additional parameters required for the random distributions (alpha, tau, gamma) are either fixed or the user needs to supply a vector that is of the required length, which is not the number of gradient locations evaluated because of the way expand() is used to replicate species parameters per set of gradient locations.

The main use-case I envisage is per species values of alpha, gamma, and tau that need to be replicated in the order that species' parameters get replicated. However, these parameters are not supplied via the usual parameters for the species so need to expanded separately.