We consider the design of Markov chain Monte Carlo (MCMC) methods for large-scale, distributed, heterogeneous compute facilities, with a focus on synthesising sample sets across multiple runs perfor
