ANSWER: NOTHING INITIALISES THE BAND ELEMENT'S BYTE 4.

Test: photomap created with the correct xieCreatePhotoReq (levels 2,2,2,
#components 1, bitonal), then SetTransport(mode=3, org=1) and PutTile(64x64,
plane_mask=1, final=1).  A write watchpoint was armed on element+0x20 (byte 4)
as soon as the element existed, with the stack control watch proving the
mechanism was live (952 control hits).

  obj=003FD7D0  band=003DBAF0  elem=003FD590  elem.b4=55
  after PutTile:                              elem.b4=55        <-- UNCHANGED
  watch hits: 1, at pc=3F4019D0 (file 0x4033A), val=0000B055
     -- that instruction is `MOVE A8,*A9(0x28)`, i.e. it stores byte 5, the
        adjacent field, not byte 4.

So across create -> SetTransport -> PutTile the element's compression byte stays
at the heap fill value 0x55.  The import path never initialises it, and the
export path later copies it verbatim into renderptr->compression_scheme
(0x1B256: MOVB *A5(0x20),*A11(0x630)), where DDX_RENDER rejects it with
"unsupported compression scheme".

That is the complete, measured causal chain from a correct client request to the
BadValue the client receives.

STATE
  import verified: element+0x28 = 0xFF white / 0x00 black, QueryResource
    comp_COUNT=1, 64x64, constrained
  export: BadValue in every configuration, from both the hand-rolled client and
    DEC's own 2.3 client built from source
  with a labelled poke setting element byte 4 = 0 (XieK_PCM), the renderer
    passes both DDX gates (CHK_610 val=00911050, PAST_610_OK) and then fails on a
    further, undocumented path
  window: unchanged in every run; white vs black captures differ only by
    Terminal Manager figures and cursor
  xshow (core X) remains the only viewer that has shown an image

STATUS.  NO client-chosen XIE image has been displayed.
