FIRST CHAIN WHERE arm1, PUT_TILE AND EXPORT ALL RETURN SUCCESS.

Key realisation from the register dump: the two checks read DIFFERENT fields, so
they are NOT mutually exclusive after all --
    arm    requires object->byte8 == 1            (DSJS at 0x6ACC2, a7 = byte8)
    import requires *(object->byte36) == 3        (dereferenced byte, 0x6BAF2)
Earlier turns concluded they conflicted only because I assumed both were byte8.

CHAIN (create w9=0x100 -> byte8=1; byte36/scratch poked to satisfy the import
gate; debug pokes labelled):

  white tile                                          black tile
    arm1      SILENT(ok)                                SILENT(ok)
    PUT_TILE  SILENT(ok)   <- import accepted            SILENT(ok)
    arm2      ERROR 8                                    ERROR 8
    EXPORT    SILENT(ok)   <- export accepted            SILENT(ok)
    RP        src_base610=00000000                       src_base610=00000000
    window    nz 512 -> 512, changed=False               changed=False
              first=888888888888888822222222...          same

So for the first time the import AND the export both report success -- but
  * src_base (rp+0x610) is STILL 0,
  * the window is unchanged, white and black identical,
  * arm2 (the second execute) still returns BadMatch.

WHAT THIS MEANS: the import being accepted does not populate the export's source
buffer.  The handoff that should fill element+0x28 (the import-element executor
0x7EAB640, which the earlier analysis tied to this path) evidently does not run,
or runs without a destination.  arm2 failing may be part of the same gap.

NEXT: instrument 0x7EAB640 (file 0x6B640) during this now-successful PUT_TILE to
confirm whether the import handoff executes at all, and if it does, what it
writes -- since that is the only remaining step between "import accepted" and
"src_base non-zero".

Status: no client-chosen XIE image yet; the window remains the stipple; xshow is
still the only viewer that has displayed an image.
