THE IMPORT HANDOFF 0x7EAB640 NOW EXECUTES (first time in this investigation).

Two more conditions were needed beyond the gate, both found by instrumenting:

  1. the lookup's mask check at 0x6BB82 (CMP A7,A11) compares *(byte36 + 8)
     against the requested band.  With band = 1 the byte at byte36+8 must have
     bit0 set:
         byte36->[0]=3, [8]=0, band=1 :  SELB_CMP a7=0 a11=1 -> SELB_BAD8
         byte36->[0]=3, [8]=1, band=1 :  SELB_CMP a7=1 a11=1 -> OK_6BBE2
  2. with that satisfied the lookup returns 0 and PUT_TILE is accepted.

  Debug pokes used: object->byte36 = scratch; scratch[0] = 3; scratch[8] = 1.
  (All pokes labelled; client reachability still unproven.)

RESULT (both tiles, identical):

  arm1      SILENT(ok)
  PUT_TILE  SILENT(ok)
  handoff   HANDOFF_ENTRY 0x6B640 -> H_84C_NOOP      <-- RUNS, then NO-OPS
  arm2      ERROR 8
  EXPORT    SILENT(ok)
  window    nz 512 -> 512, unchanged, 888888888888888822222222
  src_base  0 (RP never printed non-zero)

So the handoff executes but immediately takes its no-op branch at 0x6B84C, which
is the guard
    0x6B684 MOVE *A9(0xd8),A9,1
    0x6B688 JREQ  0x6B84C          ; band subobject == 0 -> no-op
i.e. the band subobject pointer at band+0xd8 is zero for this resource.

PROGRESS SUMMARY (each step measured, in order):
  byte8 == 1                      -> arm passes
  *(byte36) == 3                  -> gate at 0x6BAF4 passes
  *(byte36 + 8) bit0 == band      -> lookup mask check at 0x6BB82 passes
  band (word5) != 0               -> import path not skipped at 0x6A896
  -> handoff 0x6B640 finally RUNS, but band+0xd8 == 0 so it no-ops.

NEXT: make band+0xd8 non-zero (the band subobject / element record).  That is the
last gate before tile data can reach element+0x28 and hence src_base.

Status: no client-chosen XIE image yet; window unchanged; xshow (core X) remains
the only viewer that has displayed an image.
