TURN RECORD: THE rp+0x48 TEST PASSES; 0x1BA30 IS A SHARED REPORT BLOCK REACHED BY
AN UNCONDITIONAL JUMP, AND IT REPORTS WHATEVER CODE WAS ALREADY IN FRAME 0x120.

MEASURED (both id low-bytes, with instrumentation on the branch itself):
  MAP=00400B00  J884 rp=009261D0 b9=01  -> CONT_1B888  ... ERRCHAIN_1BA30  RET a8=2
  MAP=00400B01  J884 rp=00926910 b9=01  -> CONT_1B888  ... ERRCHAIN_1BA30  RET a8=2

So:
  * rp byte 9 is ALREADY 1 in both cases -- the 0x1B884 JREQ 0x1BA30 is NOT taken;
    the earlier "flag is zero" reading was wrong.
  * 0x1BA30 is nevertheless reached, from 0x1B9F2 `JR 0x0001BA30` -- an
    UNCONDITIONAL jump that is the normal continuation after
    `CALLA 0x07E5BBCA` returns 0.
  * 0x1BA30 is therefore not an error branch but a shared epilogue that:
        0x1BA30 CALLA 0x07F67F38 -> @0xa5e60
        0x1BA44 CALLA 0x07E75EA8  (called with 0xa6000 -- frees/marks the rp)
        0x1BA60 CALLA 0x07EA45FE
        0x1BA68 MOVK 0x2,A11 ; ... ; 0x1BA78 CALLA 0x07E7FE96
    i.e. it REPORTS THE CODE ALREADY HELD IN FRAME SLOT *A13(0x120), which is 2.

CONCLUSION: the 2 is decided earlier in the driver and merely reported here.
The chain to chase is whatever stores 2 into *A13(0x120) before 0x1B9F2.

STATE (unchanged):
  import verified (element+0x28 = 0xFF white / 0x00 black, comp_COUNT=1)
  export: err2 in every configuration; window never changes
  viewer: xshow (core X) is the only thing that has displayed an image

STATUS.  NO client-chosen XIE image.
