I am working on a Monte Carlo simulation which consistently runs chains in parallel (each running for a fixed amount of steps) and gathering statistics on them. I am wanting to use persistent threads for this purpose but am very confused on how to begin and how to set this up. Ideally, I’d also want to be able close the kernel after obtaining a number of samples, but I would not want to specify the number of samples in advance, as they might be very large, and might not be known in advance (i.e. based on a convergence criteria of the statistics being gathered).