Case 06 · Incident reconstruction

Reliable Long-Running Automation

Make long-running work explainable when it stalls: preserve the last healthy checkpoint, freeze evidence, apply one bounded repair, and prove integrity before completion.

Open the incident reconstruction

Primary decisionFence the stale attempt

Primary resultContinuity certificate

LimitationNo real job, prompt, session, archive, host, path, account, identity, credential, service, schedule, worker, filesystem write, or background process is shown or started. No deployed integration is claimed.

Black Box Reconstruction · synthetic

Turn a broken heartbeat into evidence, one bounded repair, and an integrity decision.

Synthetic replay · no background job · topology withheld

Run contract

Units: 8
Attempts: 2
Checkpoint: 03
Repair budget: 1

Flight recorder core

PendingNot observedfuture evidence withheld
PendingNot observedfuture evidence withheld
PendingNot observedfuture evidence withheld
PendingNot observedfuture evidence withheld

Unit 01 · verifiedUnit 02 · pendingUnit 03 · pendingUnit 04 · pendingUnit 05 · pendingUnit 06 · pendingUnit 07 · pendingUnit 08 · pending

Not yet observed

Integrity chamber

expected Units: Not yet observed
completed Units: Not yet observed
duplicate Units: Not yet observed
missing Units: Not yet observed
orphan Receipts: Not yet observed
evidence Items Expected: Not yet observed
evidence Items Observed: Not yet observed
fixture Digest Parity: Not yet observed
status: Not yet observed

Contract armed · units and repair budget fixed

Continuity Certificate

8/8 units · 0 duplicates · 0 missing · 4/4 evidence · integrity pass · network none · persistence none · external actions 0

Attempts / heartbeats: 2 / 4
Checkpoint / failure: 03 / TRANSIENT_UNIT_FAILURE
Repair budget: 1/1
Digest parity: true / pass
Digest: fx-586f3c0e

run contract
last healthy heartbeat
checkpoint 03
failure observation

Inspect evidence boundary

MaturityComposite technical evidence

EvidenceThree independent sources · synthetic composite replay

The operating problem

Completion without evidence is only a hopeful status.

Long-running automation needs a durable contract, last healthy checkpoint, frozen failure evidence, a repair budget, fenced retries, and integrity reconciliation before it can claim success.

Independent proof

Independent patterns, one fictional reconstruction

Three independently evidenced patterns → one synthetic incident reconstruction. No deployed integration artifact is claimed.

Archive continuity pattern

Technical product

Current local and recorded · 2026-08-29neutral label · no source media

Append-safe capture, exact or live-prefix verification, checkpoints, coverage and integrity state.

No private archive, message, session, path or installed topology is shown.

Bounded control pattern

Internal technical product

Recorded source verification · 2026-08-29neutral label · no source media

Declared work, bounded retries, proof requirements, fail-stop state and final reconciliation.

No real run, prompt, worker, host, account or orchestration artifact is shown.

Operational recovery pattern

Technical library

Recorded source verification · 2026-08-29neutral label · no source media

Job states, heartbeats, failure snapshots, stale detection and explicit recovery boundaries.

Dirty working-tree addition is excluded; no detached process or provider flow is invoked.

Written-first handoff

Need automation that explains what happened after the happy path?

Start with the work contract, checkpoint, heartbeat, failure evidence, retry budget, integrity counts, and the exact proof required before completion.

Open the written brief