- A completed AWS recovery campaign lasting three hours.
- 2,338 forward/recovery pairs with baseline verification and no recorded failures.
- Evidence custody for the campaign archive and SHA-256 digest.
PINCR · EVIDENCE RECORD · 2026-09-21
A completed recovery campaign. Clear boundaries.
RF completed an AWS recovery qualification campaign lasting three hours across a live PostgreSQL HA environment: 2,338 verified recovery pairs, no recorded failures.
Three hours. Six guarded lanes. Every recorded pair passed.
The campaign ran as six sequential lanes of 30 minutes, each using a fresh table, avoiding PostgreSQL limits on physical slots for dropped columns while preserving the same forward, recovery, and workflow for baseline verification.
These six lanes are endurance lanes of the recovery campaign. They are not the H1–H4 cells of the high-availability campaign, which tests a different set of faults.
| Lane | Result | Pairs | Failures |
|---|---|---|---|
| H1A | PASS | 392 | 0 |
| H1B | PASS | 385 | 0 |
| H2A | PASS | 391 | 0 |
| H2B | PASS | 386 | 0 |
| H3A | PASS | 388 | 0 |
| H3B | PASS | 396 | 0 |
| Total | PASS | 2,338 | 0 |
The supervising service exited cleanly with status 0. The campaign archive is hashed and retained in evidence custody.
02 · BOUNDARIES
What the receipts establish. Where the claim stops.
- Loss of a leader, loss of quorum, or every type of infrastructure failure. This campaign did not inject them. Leader-host loss and ambiguous remote outcomes were tested separately in the September 2026 high-availability campaign, which passed all four of its cells in a single ordered run on 1 October 2026, within the bounds that run states.
- Universal production readiness or commercial qualification.
- Any latency percentile or fault outcome not recorded in the archived receipts.
- A revision of PINCR preserved prior negative evidence.