P3-9.5 How Is the Same Event Tracked Across Outputs?¶

Section ID: P3-9.5 Version: v2026.09.20

A common ID identifies a subject, but not which calculation about it was used. If window A was compared against two baselines, window_id=A alone cannot identify the queue's evidence. Linking outputs requires the subject ID together with the report, policy, and outcome evidence actually used.

Window A and its member operations use different identifiers¶

Continue with comparison window A from P3-9.3. A groups 20 recent completed operations; it is not a single operation's event_id. Assign those operations the following fictional IDs to illustrate linkage. The two display rows group the list for readability; the underlying relationship consists of 20 window_id–event_id pairs.

Twenty operations link to one window. Overlapping windows may also contain the same operation, so record each membership. Changes to membership require a new window ID or a separate membership version. A's membership is fixed here. Do not copy its inspection outcome to every operation E001–E020.

Why A's two reports are not interchangeable¶

Retain A's −0.35 L/min difference from the preceding section and add fictional recent and baseline means of 2.20 and 2.55 L/min to expose its calculation. All times below share one time zone on 2026-09-20. The second report changes only the baseline; a newer version is not automatically more appropriate.

Means and differences use L/min, with equal weights for per-operation late-period means. 2.20−2.55=−0.35 and 2.20−2.30=−0.10. A's recent observations and membership did not change. The difference shifted by 0.25 L/min because the baseline changed, not as evidence of recovery.

base-v1 and base-v2 must resolve to baseline records specifying reference periods, equipment and operating conditions, included operations, and aggregation definitions. Names without retained definitions cannot reproduce calculations. These summaries illustrate differences; they do not establish which baseline better fits the task.

The queue identifies the report actually used¶

Allocation Q-01, created at 10:30, uses queue-v1 from P3-8.5 and capacity two. B retains difference −0.35 and safety category no; C retains −0.18 and yes. A is yes and uses REP-A-1's −0.35, yielding A→C→B and selected entries A and C.

This is an excerpt containing only A's selected queue row. Preserve C's queue row and original candidate B separately. Without source_report_id, A's two reports are ambiguous. Attaching the latest report would insert REP-A-2, created at 11:00, into a decision made at 10:30.

A new allocation using REP-A-2 would rank C's 0.18 before A's 0.10 within the yes group, giving C→A→B. Record a new allocation ID and time rather than overwriting Q-01's historical evidence. Times help cross-check chronology, but reports may share timestamps, so they cannot replace report IDs and explicit references.

Trace an outcome candidate back to its inspection note¶

Link A's inspection outcome one from P3-9.3 in the same way. The identifiers and times below are fictional additions for this tracing exercise. The note reference illustrates source identity; it is not an actual file.

The candidate outcome one comes from the inspection note, the difference from REP-A-1, and the rank from Q-01's policy application. Their sources differ. inspection-v1 must resolve to the definition of the issue and inspection scope. Do not use a result note written at 12:00 as information already available for allocation at 10:30.

Tracking Events Across Outputs with Shared Identifiers¶

flowchart TD
    A["Outcome candidate: A = 1"] --> B["Q-01 / queue-v1"]
    B --> C["REP-A-1 / base-v1"]
    A --> D["NOTE-A-1 / check-A"]
    C --> E["Window A"]
    E --> F["Members: E001 to E020"]

For a queue row, follow Q-01→REP-A-1→base-v1 for comparison evidence, and the outcome candidate's NOTE-A-1→check-A for inspection evidence. Resolve member operations separately through A→E001–E020. Collapsing everything into one event_id loses the distinct units of window, calculation version, allocation, and inspection.

Repair the wrong report link¶

Someone attaches REP-A-2's −0.10 to A in Q-01 because it is “the latest value,” then copies confirmed_issue=1 to all E001–E020. Identify both errors and restore the correct links.

Explanation: Q-01 specifies source_report_id=REP-A-1, so its historical difference is −0.35 and its baseline is base-v1. Preserve REP-A-2 separately and create a new allocation if needed. Outcome one belongs to the inspection associated with A, not automatically to each operation. Individual-operation labels require evidence at that unit.

Checklist¶

• Can you trace a queue row beyond the subject ID to its actual report and baseline version?
• Can you link an outcome candidate to its inspection note, reviewer, and definition version without mixing in information unavailable at the time?
• Can you record window and operation IDs separately without automatically propagating outcomes across units?

Sources and references¶

• W3C, PROV-Overview — Reference for tracing entities, processing activities, derivations, and versions. Numbers, IDs, and times are fictional examples constructed for this book. Accessed: 2026-09-20.