PUT_TILE "SUCCESS" IS A NO-OP: THE IMPORT HANDOFF NEVER RUNS.

Instrumented the handoff (file 0x6B640) and slot 19's own path during a
PUT_TILE that now returns success (byte8=1, byte36 pointed at a byte == 3):

  handle breakpoints set: HANDOFF_ENTRY 0x6B640, H_684, H_688, H_6EA, H_84C,
                          H_8D6, plus S19RET.
  result trace:           S19RET a8=00000000        (lookup returned 0 = ok)
                          ... and NOTHING from the handoff.

Second run, tracing slot 19's own landmarks:

  S19RET a8=00000000        (0x6A7B4: lookup succeeded)
  S19_7BE   0x6A7BE
  S19_850   0x6A850
  S19_ERR_EXIT 0x6AA4A      <-- slot 19 exits here
  (S19_9A2, the CALLA 0x7EAB640 handoff call, is NEVER reached)

So with the gate satisfied, slot 19 accepts the request, walks to 0x6A850, and
then takes the exit at 0x6AA4A without ever calling the import-element executor
0x7EAB640.  A8 is 0, so the client sees "success" -- but no tile data is
transferred, and element+0x28 is never written.  That is exactly why src_base
(rp+0x610) stays 00000000 even when arm1, PUT_TILE and EXPORT all report success.

RECAP OF THE NOW-COMPLETE GATE PICTURE (all measured):
  arm    requires object->byte8 == 1                   (DSJS 0x6ACC2, a7 = byte8)
  import requires *(object->byte36) == 3               (deref at 0x6BAF2)
  -> these are independent, so a byte8=1 resource with byte36 -> byte==3 passes
     BOTH and PUT_TILE returns 0.
  BUT passing the gate only reaches the exit at 0x6AA4A; the actual import
  (0x7EAB640) sits on a different branch that is not taken.

NEXT: read the code between 0x6A850 and 0x6AA4A to find the branch that bypasses
0x6A9A2, and what it tests.  That branch is now the single remaining obstacle
between "gate passed" and "tile actually imported".

Status: no client-chosen XIE image; src_base 0; window unchanged; xshow is the
only viewer that has displayed an image.
