Ah, you mean like Object>>doOnlyOnce: and Object>>reArmOneShot? Here
is a pattern which I often use (in terms of those two methods):
normalMethod
 self doOnlyOnce:[
  [self halt
  self tryMyWeirdMethod] ensure:[self reArmOneShot].
 ].
What does this do? The first time the system gets there, it'll halt
and I can (from the debugger) step into my new version of the method
and walk through it, programming happily along. If I get something
wrong (or make a mistake otherwise) I just close the debugger which
only then will get me a new debugger for continuing to program.
When I am mostly confident with the method, I simply remove the "self
halt", and leave it running for the entire system. If something goes
wrong, it'll fall back to the original implementation during the debug
session.
I found this to be the most useful technique for debugging UI problems.