Yes, but only if the acceptance record points to the exact rule, parser, configuration and bounded outcome evidence used at the time. A job ID and a success hash alone cannot tell a reviewer why last month’s authorized proxy job was accepted.

This guide is for the owner of an authorized recurring public-data, regional-check, or advertising-verification workflow. It supplies a minimal decision ledger that lets a later reviewer reconstruct one acceptance decision without retaining a request body, response body, credential, private target, or a made-up retention period. Reconstruction is not replay, proof that the source was true, proof that the job really ran, or proof that a provider performed well.

Record the decision’s dependencies, not a success slogan

Write one row when the decision is made. Keep stable references to the authorized task class and the person responsible, then link the decision to immutable or otherwise retrievable definitions: the acceptance rule version, parser version, effective configuration version, and a redacted outcome summary. The W3C PROV model describes provenance as information about entities, activities and responsible agents; the practical ledger below uses that idea narrowly for a job decision. OpenLineage likewise treats a job as an evolving process whose changes are captured during runs. Neither specification verifies your source, authorization, implementation, or result.

Ledger fieldSynthetic exampleWhy a later reviewer needs itWhat it cannot establish
Decision and scopeJOB-SYN-001; AUTH-SYN-24; synthetic reviewerIdentifies the bounded accepted job, authority reference and accountable reviewer.That authorization was valid or the target was permitted.
Rule definition and fingerprintaccept-v3: SKU and currency required; validator outcome VALID; SHA-256 of canonical definitionShows the archived criterion and detects a later mutation under the same label.That the criterion was appropriate for every task or its history is authentic.
Parser definition and fingerprintparser-7: emits SKU and currency; SHA-256 of canonical definitionConnects the claimed fields to the parser contract used then and detects mutation.That extraction was semantically correct against a real source.
Effective configuration and fingerprintroute-policy-2: approved read route class; SHA-256 of canonical definitionSeparates a deployed/effective setting reference from a repository filename and detects mutation.That a real process loaded it or used a particular exit.
Outcome and evidence referencenormalized, redacted required-field summary; SHA-256 digest; EVID-SYN-001Detects a mismatch against the declared synthetic summary without exposing a payload, and names the controlled evidence reference.Authenticity, sender identity, source truth, or a complete event history.

Also record the decision result, time and owner. Use a controlled evidence location only when its access and purpose are approved. Do not turn this ledger into a copied log. A digest is a consistency check over supplied bytes; it is not provenance, an attestation, a signature, authorization, or authenticity. If the definition or controlled evidence is no longer available under the applicable policy, record that the decision cannot be reconstructed at that level. Do not retain sensitive payloads merely to improve a future explanation, and do not invent a retention duration here.

Make the old-versus-current comparison inspectable

In the fixture, the archived definitions are accept-v3 = {"required":["sku","currency"],"outcome":"VALID"}, parser-7 = {"fields":["sku","currency"]}, and route-policy-2 = {"route_class":"approved_read"}. The normalized redacted outcome is {"job_id":"JOB-SYN-001","required_fields":["sku","currency"],"validator_outcome":"VALID"}. The later accept-v4 changes only the required fields to ["sku","currency","tax_basis"]. A reviewer can therefore reproduce the delta without a source payload: the archived rule accepts the declared outcome, while the current rule would not. The fixture also pins a SHA-256 fingerprint for each archived definition; a changed definition behind an unchanged version label returns HOLD. That detects inconsistency in supplied objects, not authentic history.

Ask the reconstruction question in the right order

  1. Find the decision row and confirm it names one authorized scope, owner, rule version, parser version, configuration version and outcome claim.
  2. Retrieve the referenced definitions through the approved internal path. If any definition is absent, ambiguous, altered without a history, or outside the review scope, return HOLD.
  3. Check that the rule’s required fields match the parser’s declared output contract, and that the outcome claim meets that rule.
  4. Recalculate any digest only against the approved, redacted summary. A mismatch means HOLD; it does not identify a culprit.
  5. Report a narrow result: “the acceptance rationale was reconstructed from these references” or “it could not be reconstructed.” Keep source accuracy, route observation, replay eligibility and authorization as separate questions.

Do not “fill in” a missing old version by running today’s parser or current configuration. That creates a new observation or test, not evidence about the old decision. Historical correction and replay have their own authorization, input-retention and representation requirements; see the parser repair and replay plan. A configuration change needs its own impact and result checks; see configuration-change revalidation.

A small synthetic acceptance and six fail-closed cases

All labels and outcomes below are fictional. The positive row is reconstructible because every referenced definition exists and agrees. The other rows stay on hold; they do not become rejected business results, and they do not justify replaying a request.

CaseLedger referenceResultReason
Positiveaccept-v3, parser-7, route-policy-2, matching summary digestRECONSTRUCTEDThe declared field contract, outcome and route class agree.
Missing parserparser-absentHOLDNo definition explains how the claimed fields were produced.
Missing configurationroute-policy-absentHOLDA route label cannot substitute for its versioned definition.
Outcome conflictaccept-v3 plus INVALIDHOLDThe claim does not meet the referenced acceptance rule.
Digest conflictKnown versions, mismatched summary digestHOLDThe supplied summary is not the one named by the ledger.
Malformed ledgerNo object suppliedHOLDAn absent or malformed entry cannot explain a decision.
Definition mutationaccept-v3 label, changed definition fingerprintHOLDThe supplied archived rule no longer matches the pinned definition.

The worked record was accepted under archived accept-v3, which required SKU and currency. A later accept-v4 additionally requires tax_basis. The old redacted outcome lacks that field, so it is NOT_ACCEPTED_UNDER_CURRENT_V4. That version-linked delta explains the difference; it does not rewrite the historical acceptance, authorize a replay, or decide whether a source value was true.

Run the offline fixture before adapting the pattern

Download the Python 3.9+ standard-library fixture and run python3 proxy-job-decision-provenance.py. Its exact output is:

positive accepted decision: RECONSTRUCTED — accepted under archived accept-v3 with parser-7 and route-policy-2
version delta: ACCEPTED_UNDER_ARCHIVED_V3; NOT_ACCEPTED_UNDER_CURRENT_V4
negative missing parser definition: HOLD — missing parser definition
negative missing configuration definition: HOLD — missing configuration definition
negative outcome claim: HOLD — outcome claim does not satisfy the referenced rule
negative mismatched outcome digest: HOLD — outcome-summary digest does not match the declared synthetic summary
negative malformed ledger: HOLD — ledger entry is not an object
negative changed archived rule definition: HOLD — referenced rule definition differs from its archived fingerprint
PASS: synthetic fixture only; no network, files, secrets, or source payloads.

The fixture canonicalizes synthetic JSON and uses hashlib.sha256; Python documents that a digest is computed over supplied bytes. It has no network, reads no environment, files, secrets, logs, payloads or production configuration, and changes nothing. It checks only that its fictional ledger is internally consistent. Owner, time and evidence-location fields are checked for a non-empty string only; the script does not validate the date format, event chronology, reviewer identity or whether the evidence reference resolves. A human must check those items through the approved records before claiming a historical explanation. It cannot verify a contract, source statement, collection permission, run-time loading, artifact history, signature, authenticity, data completeness, business outcome, IPHTML capability, or a suitable retention policy.

Choose a bounded next step

For an authorized workflow with a real location-specific observation gap, first check whether an official API, licensed dataset, direct approved access, or manual process answers the task. If a residential context is still relevant, review IPHTML residential proxy information. For a scoped discussion, use the contact route with the authorized task class, required region, rule/config version references, decision deadline and exclusions—never credentials, raw responses, private targets, or customer data. This method contains no measured IPHTML performance, price, availability, capacity, support, demand, payment or renewal claim.

Sources checked November 22, 2026: W3C’s PROV-DM for provenance concepts; OpenLineage job facets for evolving-job context; and Python’s hashlib and json documentation for the offline fixture. They do not prove a real job, source truth, retained evidence, authorization, product performance, customer demand, payment or renewal. All examples are synthetic.