The standard we test others against, applied to ourselves. Errors are not deleted here — they are dated.
In audit work, credibility is binary. If we cover up our own validation errors or documentation mismatches, we lose the right to check anyone else's models. This registry documents every mistake caught in our pipelines or pre-registration plans.
The pre-registration froze the F4 hypothesis around a 76.8 MWh telemetry threshold for April 1, 2026. That number does not exist in the raw data (actual max_soc: 102.57 MWh; soc: 73.77 MWh). The scoping error passed the L0 gate into the frozen protocol, voiding the pre-registered F4 test.
The error is documented in audits/US-TX-BATC-001/failures.md. The F4 verdict is Deferred pending official ERCOT column schemas — no post-hoc reframing was used to rescue the hypothesis. The rules stayed frozen; the mistake stays visible.