I have recently noticed that my tests take longer to run and it appears to be due to the latest version of rstan (2.19.2). To verify this, I ran some of our Stan code with versions 2.18.1 and 2.19.2 for 10 different seeds, and observed that the latest version was consistently running around 10% slower than the previous version with 2 chains and 200 iterations. To minimise measurement noise I ran the calculations from safe mode in Windows. My R code is below along with two CSV attachments containing data, which should provide a runnable example. Has anyone else noticed a difference in performance?

```
data <- list(C = 3L, R = 380L, RS = 3040L, V = 13,
prior_mean = c(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0),
prior_sd = c(5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5))
data$X <- array(c(as.matrix(read.csv("xdata.csv"))), dim = c(3040, 3, 13))
data$Y <- read.csv("ydata.csv")[[1]]
data$S <- rep(8, 380)
options(mc.cores=parallel::detectCores())
library(rstan)
stan.model <- stan_model("hb.stan")
result <- sampling(stan.model, data = data,
chains = 2, iter = 200,
control = list(max_treedepth = 10,
adapt_delta = 0.8),
init = function () list(L_omega = diag(13)))
```