TURN RECORD: THE FINALIZER RETURNS 0, SO THE err2 IS STILL FURTHER OUT.

Instrumented the render driver's finalizer path (0x1B842 -> CALLA 0x07E79F32)
and the helper's own BadValue site (file 0x39F32):

  0x39F54  A11 = rp
  0x39F58  A11 = *A11(0x20)      ; rp byte 4
  0x39F5C  JRNE 0x39F64          ; non-zero -> continue
  0x39F5E  MOVK 0x2,A8           ; else -> return 2

MEASURED on a live export:
  RP20 rp=008FD510 byte4=01 -> forcing 1
  RP20_OK_CONTINUE
  AFTER_FINALIZE a8=00000000     ; the finalizer SUCCEEDS

So:
  * rp byte 4 was already 1 -- that BadValue site is NOT the one firing;
  * the finalizer returns 0 (success);
  * the client still receives err2 and the window still does not change
    (white vs black AE = 2040 px, i.e. the Terminal Manager's own figures and
     the cursor, not the 64x64 window).

The err2 therefore originates even later in the chain than 0x1B842 -- after a
successful finalize -- or in the export handler's post-render cleanup
(0x630B0 CALLA 0x07F63B00) before it reaches 0x630E8 (which would return 0).

STATE OF THE OBJECTIVE (unchanged)
  import: verified in the server (element+0x28 = 0xFF white / 0x00 black)
  src_base: non-zero (008FD510 / 00915FD0)
  export: always err2; window never changes; white and black captures identical
          apart from the Terminal Manager's own figures and the cursor
  viewer: xshow (core X) is the only thing that has ever displayed an image

STATUS.  NO client-chosen XIE image.
