SLOT 44 OPERAND REQUIREMENT FOUND -- and the first element-type write ever seen.

Static requirement, from the handler (file 0x65CAA, 5 words):
  65CFA MOVE *A9(0x60),A7,0x1     ; payload word 2
  65CFE JREQ 0x65D06              ; word2 == 0 -> other path
  65D00 MOVE *A9(0x80),A7,0x1     ; payload word 3
  65D04 JRNE 0x65D2E              ; word3 != 0 -> main path
So BOTH word2 and word3 must be non-zero.  Every earlier sweep of slot 44 used
word3 = 0, which is why it always came back BadValue.

MEASURED (valid ids, two live resources A/B, SET_OP_DEFAULTS(1) first):

  [A,B,0,0]  both zero  -> reply        render=0  types=-
  [A,B,1,0]  w3 zero    -> BadValue(2)  render=0  types=-
  [A,B,0,1]  w2 zero    -> BadValue(2)  render=0  types=-
  [A,B,1,1]             -> BadValue(2)  render=0  types=['TYPE2']   <-- !
  [A,A,1,1]             -> BadValue(2)  render=0  types=['TYPE2']   <-- !
  [A,B,3,3]             -> BadValue(2)  render=0  types=['TYPE2']   <-- !

FIRST ELEMENT TYPE EVER WRITTEN: the type-2 store (file 0x1B362,
MOVK 0x2,A11 ; MOVB *A11,*A5(0x58)) now executes.  In every previous
configuration, including the full export cycle, NO type store ever fired.

So slot 44 with two non-zero trailing fields DOES reach the element-add path and
selects element type 2.  It then returns BadValue -- so a later check rejects the
request -- and the render call at 0x65F54 did not fire.

CAVEAT RECORDED HONESTLY: I armed a breakpoint at the 0x65F54 call site and it
did not fire even though the type write (which lives inside the callee 0x1B152)
did.  Either the call is reached by a different route than the site I targeted,
or the breakpoint was dropped despite only 7 being armed.  I did not resolve
which, so the exact call path from slot 44 into 0x1B152 is not established.

NEXT: bisect the remainder of slot 44 after the type write (0x65D2E onward) to
find the check that still returns BadValue, since that is now the only obstacle
between us and a constructed element.
