I had previously checked out the re pragma's debug functions but I didn't find them very useful. For example, either of the regexes in the OP generates:

synthetic stclass "ANYOF[a-z][]".
Final program:
   1: BOL (2)
   2: OPEN1 (4)
   4:   PLUS (16)
   5:     ANYOF[a-z][] (0)
  16: CLOSE1 (18)
  18: STAR (20)
  19:   DIGIT (0)
  20: EOL (21)
  21: END (0)
floating ""$ at 1..2147483647 (checking floating) stclass ANYOF[a-z][]
+ anchored(BOL) minlen 1
[download]

It's heartening that the output for both of them is identical, as I'd expect, but I don't think parsing this output in order to reconstruct the source regex will be significantly easier than modifying the source regex directly :-(

Probably not easier, but possibly less ambiguous.

I wonder if constant regex's are compiled once during the compile phase. If so, would be useful if Deparse reconstructed the regex.

(Currently reading and posting from my tablet, so will try to remember to try to test this, later.)