SLOT 43 IS A RENDER REQUEST -- correction to the earlier "printf/describe" label.

Instrumented path for [A,A,0,0,64,64,0,0,0] on a live class-2 resource
(9 breakpoints, generated addresses, valid ids, isolated connection):

  S43_LOOKUP 0x65646  (lookup returned 0 -- success)
  S43_TAIL_A 0x65A3A
  S43_TAIL_B 0x65B78
  S43_SET3   0x65BD0   MOVK 0x3,A5
  S43_STORE3 0x65BD2   MOVB A5,*A8        <-- writes the value 3 into the new block
  S43_RENDER 0x65BF8   CALLA 0x07E5B152   <-- the SAME render/export driver slot 21 uses
  (S43_EXIT0 0x65C9E never reached)

So slot 43:
  1. looks up the resource (word 0 must be a class-2 image resource; word 0 = photoflo
     gives 0xE, word 0 = image gives 0),
  2. allocates and fills a process block,
  3. sets a byte of it to 3 (0x65BD2),
  4. invokes the render driver 0xE5B152 = file 0x1B152, the same function slot 21's
     export calls,
  5. never reaches its own `SUB A8,A8` return at 0x65C9E.

The client still receives BadValue (2), and the two MOVK 0x2 sites (0x658C0 and
0x6598C) do NOT fire.  So the 2 is produced by a callee inside the tail -- the
candidates in that range are 0x7E5B152 (render driver), 0x7F16CA2 (table insert),
0x7E7FE96 and 0x7F63B00 (free).

CONSEQUENCE FOR EARLIER CLAIMS:
 * "Slot 43 is a printf/describe request" was WRONG.  It formats doubles for a
   report, but its substantive action is to build a process block and call the
   render driver -- i.e. it is a second, independent entry into the render path.
 * It is still not an image SOURCE (it never writes resource+0x140 or rp+0x610 as
   a buffer), but it is a render TRIGGER and was misclassified.

NEXT: bisect inside the tail (0x65C02..0x65C9E) and inside 0x1B152's return path to
find what returns 2, since an accepted slot 43 would drive the render driver
directly -- the closest any request has come to the export path.
