TURN RECORD: NO REQUEST WRITES THE ELEMENT RECORD, AND THE DEC XIE SERVER
DOCUMENT WAS FOUND (it confirms the missing request shape).

A) WATCHPOINTS ON THE ELEMENT RECORD -- ZERO WRITES FROM 33/40/43/44
   Live write watchpoints (proven: stack control fired 3052 times) on
   band_array+0xd8 and the whole element block (+0x800 bits), issuing:
      slot33 [A,B,0]   -> err2   elem_changed=False  writes=0
      slot40 [A,B,0]   -> err2   elem_changed=False  writes=0
      slot44 [A,B,1,1] -> err2   elem_changed=False  writes=0
      slot43 len 3/5   -> err16 (BadLength)
   Full slot sweep at each handler's REQUIRED length, watching the photoflo's
   p44 block: no slot produced a pipeline element.

B) SLOT 43 IS THE ONE THAT REACHES THE RENDERER -- AND STOPS THERE
   Slot 43 requires length EXACTLY 10 (0x65614: MOVE *A9(0x10) ; SUBK 0xa ;
   JREQ).  With len=10 and two valid ids:
      reply: err2 (BadValue)
      RP breakpoint at 0x1B296 FIRES:  RP a11=00291AF0 src_base610=00000000
   i.e. it enters the export/render driver, finds no elements, and returns
   BadValue.  src_base stays 0.  TYPE1/TYPE2 (the element-type stores) never fire.

C) SLOT 18 REJECTS EVEN A PHOTOFLO
   Slot 18 is the only large handler with a MINIMUM length (>=4) and it does
   touch *A7(0xd8) and write element+0x20.  Tested against a class-2 resource
   and against a class-1 photoflo:
      err10 (BadAccess) with bad = the resource id, in both cases.
   So it finds the resource and refuses access; its accepted shape is unknown.

D) THE MISSING DOCUMENT WAS FOUND
   xie/docs/dec/xieSIarch.txt  (fetched from
   https://ftp5.gwdg.de/pub/xfree86/XFree86/4.2.0/doc/xieSIarch.TXT)
   147 occurrences of "photoflo" -- this is the DEC XIE Server Interpretation
   architecture document, i.e. the right lineage (the X Consortium specs in
   xie/docs/x11r4 have ZERO).

   What it settles:
     * XIE defines six resources: ColorList, LUT, Photoflo, Photomap,
       Photospace, ROI.
     * 7.2.3 Photomap: stores a complete image; populated by an ExportPhotomap
       element; attributes inherited from that element.
     * 7.2.6 Photoflo (flo-def): "the number of elements received from the
       client" and "an array of pointers to client elements in PHOTOTAG ORDER",
       plus a definition DAG and an optimized DAG.
     * 7.3 Photo Element (pe-def): "the element phototag number received from
       the client" and "the element TYPE received from the client"; class
       Booleans import/process/export; input connections carry the phototag of
       the source element.
     * 8.3.3.1 ExecuteImmediate / 8.3.4.1 Create: "Create a flo-def, copy the
       SUPPLIED LIST OF ELEMENTS to the flo-def while verifying that the list
       is a DAG".
     * 8.3.4.2 Modify replaces element PARAMETERS only; 8.3.4.3 Redefine
       destroys and recreates under the same id.

   So elements are NOT added one at a time by scale/crop/rotate requests: the
   whole element list arrives in ONE photoflo create/redefine request, each
   element carrying {phototag, type, params}, wired by phototag.

E) WHAT THIS MEANS FOR THE REMAINING WORK
   The create path we drive (slot 8, op=1/2) checks only length >= 3 and has no
   element loop, so it is not the photoflo-create request that carries elements.
   The request that does carry them is still unidentified, and this document --
   being a SERVER ARCHITECTURE description -- deliberately does not give the
   wire encoding of that list (phototag/type/param field order), which is what
   a request has to match.

STATUS.  No client-chosen XIE image.  src_base (rp+0x610) 0 in every run; the
window is unchanged from the no-export baseline; xshow (core X) is still the only
viewer that has displayed an image.
