This development has parallels to tuning parameters in the Montecarlo world, specifically using HCMC+NUTS one does not have to worry about finding the complete distribution and you get warned when the distribution is misbehaved[0].

HCMC+NUTS not only does better exploration of the posterior but can help you select things like step sizes, acceptance rate, simulation time(See the NUTS Sampler[1]) much more easily.

[0] http://mc-stan.org/misc/warnings.html#runtime-warnings [1] http://mc-stan.org/workshops/ASA2016/day-1.pdf