compare data,table with polars

[tdhock]

for multi-threaded operations, the number of threads used by polars must be set as an option, before loading polars.
pola-rs/r-polars#720 says

#' - Set the `polars.limit_max_threads` option to `FALSE` with
#'   the [options()] function. Same as setting the `POLARS_MAX_THREADS` environment
#'   variable, this option must be set before loading the package.
...
    Sys.setenv(POLARS_MAX_THREADS = 2)
...
pl$threadpool_size() 

in contrast data.table can change number of threads interactively (without having to restart R) using setDTthreads().
Are there situations where a different number of threads is optimal?
say for operation A, 2 threads would be best, but for operation B, 4 threads would be best? in that case data.table could have an advantage. @DorisAmoakohene can you please investigate?
how can we determine optimal number of threads for polars?

• run atime, save results to disk, quit R, restart R with different number of threads, etc.
• is it possible to use atime_versions on different versions of polars?