using future_apply inside a future

[comicfans]

since fork is unstable in RStudio, I came to another solution, first using multi-session to create a future from main R session, and use multicore to fork in the future expr body(I've modified future to make supportsMulticore return TRUE from none-interactive R), so future_lapply still use fork

plan(multisession)
res = future({
   plan(multicore)
   options(mc.cores = 4)
   options(future.fork.enable=T)
   future_lapply(1:100, function(i){
        return(length(BIG_GLOBAL_VAR))
   })
})

lapply.res = value(res)

I hope such setup should makes big global variable access faster ( should be copied only once) and worker cheaper. with some simple test seems it do complete faster than directly call future_lapply with multisession

my question is : does such setup have some implicit problem ? does future/future.apply expected to work correctly in such setup ? I tried to use such setup with progressr, found that the progressor object can not work, for example

plan(multisession)
with_progress({
  p = progressor(100)
  future({ 
   plan(multicore)
   options(mc.cores = 4)
   options(future.fork.enable=T)
     future_lapply(1:100, function(i){
        p("")
    })
  })
})

is this progressr limitation, or future/future.apply should not be used in such way ? Thanks for advise