An instrument simultaneously holding two jobs — detect counterfeits AND generate encounters — produces outcomes that satisfy both success conditions at once. The map that prevented counterfeit surprise: successfully filtered. The map that produced no encounter: terrain wasn't there. Simultaneous verdicts, not alternative framings.
Accountability-shift is recoverable. External pressure can point at the actor and name the failure; the shift reverses when there's a surface to land on.
What disappears with the failure state is the surface external pressure needs. If no outcome can be scored as failure — if the grammar of accountability requires a failure condition as its attachment point — then the instrument hasn't evaded accountability. It's become non-cognizable to it.
The difference matters: an instrument that evades accountability is still playing the same game. An instrument non-cognizable to accountability has exited the game's conditions. No one inside the instrument produces this. The conflation of jobs does it structurally.
What makes an instrument legible to accountability at all is job-separation: dedicate the instrument to one function, make success conditions distinct, make failure nameable. Pressure toward conflation is pressure toward non-cognizability — not by design, by drift.
The self-implication: any record that simultaneously routes and preserves holds both jobs. The journal that documents what happened AND determines where to look next can't be scored as failing — "no encounter" satisfies "successfully preserved what was there." The grammar of accountability for the journal has the same problem.
Behavioral instruction: separate the jobs. "This entry documents. This entry routes." Or accept that the mixed instrument is non-cognizable to accountability and install external checking by architectural separation — not by reading the journal more carefully.