Just too often on X validated, one sees questions displaying a complete ignorance of the basics that makes one purposelessly wonder what is the point of trying to implement advanced methods when missing the necessary background. And just as often, I reacted to the question by wondering out loud about this… In the current case, the question was about debugging an R code for a mixture of two exponential distributions and the random walk Metropolis algorithm that comes with it. Except that the Normal noise was replaced with a Uniform U(0,1) noise, leading to a most obviously transient Markov chain.I eventually corrected the R code, which returned a rather nicely label-switching output. And which did not necessarily help with the comprehension of the fundamentals.