Why is sampling sometimes slow?

Questions v5
6

@bob-carpenter: If I understand your explanation—and I may not—you imply that differences in scale for RVs should be avoided. For example, the following model may give sampling problems:

with pm.Model():
x = pm.Normal("x", mu=0, sigma=1.0)
y = pm.Normal("y", mu=1_000_000, sigma=1_000_000_000)

Is that correct? Is there a need to noncenter even when there is nothing hierarchical?