Three Fences Any Λ Mechanism Must Pass: Identity Channels, Continua, and Dial Audits

The cosmological constant problem is not short of proposed mechanisms; it is short of cheap ways to reject them. This paper states three frame-independent necessary conditions — fences — that any mechanism claiming to predict the observed vacuum energy density ρ_Λ must pass, distilled from a completed obstruction census inside Einstein–Cartan bounce cosmology [1], where each fence arose as the kill mechanism of an entire construction class. Fence 1 (the identity-channel test): if a mechanism's predicted residue, after its own algebra is carried through, is a function of present-day quantities alone, it contains no early-universe content, whatever its intermediate steps invoke. Fence 2 (the continuum test): if a mechanism's solution set is a continuous family containing the observed value, the mechanism has not predicted the value; selection requires a discrete fixed set at the answer, for reasons internal to the mechanism. Fence 3 (the dial audit): count the mechanism's free continuous parameters, disguised ones included; a landing that needs any of them set to the answer is a reading of a dial, not a result. We then run three named external mechanism families through the fences as a calibration. Vacuum-energy sequestering fails all three: it cancels genuinely, but its residue is delivered by a global spacetime average rather than computed, ranges over a continuum with the history, and rests on global-sector scales of its own. Relaxation proposals (Abbott-type scanning and its descendants) fail all three, the dial audit hardest: a populated universe stalling near zero requires the stall structure set as a cancellation condition — Weinberg's empty-universe objection, restated as an audit. Unimodular variants fail the continuum test maximally — an integration constant admits every value — while recording exactly one dial, undisguised; their value stays open by declaration, which this paper grades as the honest best case available to a non-predictive framework, and a different thing from a pass. No fence here delivers the value of Λ, here or anywhere. The claim is a grading instrument: three necessary conditions, each cheap to apply, that a proposed prediction must clear before its heavier machinery is owed a reading.