Avoid loading samples for each iteration in variables loop inside run.sensitivity.analysis #3859

While reviewing the implementation of run.sensitivity.analysis, I noticed that samples.Rdata and sensitivity.samples.Rdata are loaded inside the loop over variables and this means the files are read from disk on every iteration:

  fname <- file.path(settings$outdir, "samples.Rdata")
  samples <- new.env()
  load(fname, envir = samples)

  fname <- sensitivity.filename(...)
  if (file.exists(fname)) {
    load(fname, envir = samples)
  }

  ...
}

Since these files do not depend on variable, they appear to be identical for each iteration, the current implementation performs repeated disk I/O that could be avoided.
I thnk Instead of loading the files during every iteration, they could be loaded once before the loop and reused during the iterations, also I think we have 2 options to refactor this:

First solution: Load the .Rdata files once before the loop using load_local() function which converts the loaded environment to a list, then Inside the loop samples <- samples_base
Since lists in R use copy-on-modify semantics, a copy would only occur when modifications are made.

Second solution: we could keep the current environment structure but load the files only once,
keep the same load logic before the loop

samples_base <- new.env()
load(samples_file, envir = samples_base)

Then inside the loop create a fresh environment copy for each iterationsamples <- rlang::env_clone(samples_base)
I think both approaches would avoid repeated disk reads overhead inside the loop.

But Before proposing a change, I wanted to ask is there a reason the files are reloaded on every iteration?