To process a large number of dirs and subdirs with subdirs... one could try threads to see if this speed up things.

How could this first shot be improved?

Try example e.g. with
======
tclkitsh xxx.tcl c:/windows
======

======
lappend auto_path ./
package require Thread
set tp [tpool::create -minworkers 10 -maxworkers 200]

foreach dir [glob -nocomplain -type d -dir [lindex $argv 0] -- *] {
   lappend res $dir
   lappend threadList [tpool::post $tp [format {
      if {![catch {glob -nocomplain -type d -dir %s -- *} res]} {
         return $res
      }
   } $dir]]
}
while {[llength $threadList]} {
   set completedThreads [tpool::wait $tp $threadList threadList]
   foreach t $completedThreads {
      set dirs [tpool::get $tp $t]
      if {[llength $dirs]} {
         lappend res {*}$dirs
         foreach dir $dirs {
            lappend threadList [tpool::post $tp [format {
               if {![catch {glob -nocomplain -type d -dir %s -- *} res]} {
                  return $res
               }
            } $dir]]
         }
      }
   }
}

tpool::release $tp
puts [join [lsort $res] \n]
======
----
<<categories>>Enter Category Here