Repeated export and partial-import recovery — working review
WORKING NOTES — review decisions belong in your controlled records.
Guide: https://seal.run/integrations/workflows/chromatography-data-integration

Use a repeated delivery, a supported later evaluation and an interrupted import from the same installed source. Account for delivery attempts separately from scientific results.

1. Source identity contract
Name the source installation, analysis, evaluation/revision and row identities. Explain how the installed export distinguishes a re-export, reprocessing and an independently acquired result. Record any identity gaps.
Your evidence and observations:
[Not recorded]

2. Delivery and content evidence
List retained delivery files and original-byte fingerprints with the algorithm used. If a normalized-content comparison is used, identify its exclusions. Record same-identity/different-content cases separately from harmless repeat delivery.
Your evidence and observations:
[Not recorded]

3. Expected-to-received row ledger
List every expected source result identity and its destination record or unresolved outcome. Include rejected/excluded rows with reasons and the mapping revision used. Count parsed, accepted and unresolved rows separately.
Your evidence and observations:
[Not recorded]

4. Interrupted and concurrent delivery outcomes
Record which rows existed before replay and which exact targets were queried before creation. Test concurrent attempts as well as sequential replay. Identify the actual uniqueness/serialization mechanism; a text field alone does not make creation atomic.
Your evidence and observations:
[Not recorded]

5. Reconciliation and result selection
State whether the intended population is completely reconciled and identify conflicts or missing outcomes with owners. Record the reviewer’s selected evaluation separately from the arrival of the latest file.
Your evidence and observations:
[Not recorded]

EVIDENCE NEEDED TO FINISH
Every intended result is accounted for by identity, all delivery attempts remain attributable and retry does not create another scientific population. Conflicting content is resolved explicitly; import completion and analytical result selection have separate decisions.