E.18.3:4.0 - Ordinary start and conditional formal recovery

Begin with the ordinary branch: name the concrete thing being transformed, mark two recognizable places or states, state the proposed connection or guard in domain language, and ask which continuation depends on it. Return a provisional explanation that either answers the question or names the missing relation, fact, or constraint. If that is sufficient, stop; neither the explanation nor the flow card must first be constituted as an episteme, position mapping, or selected structure.

Use the numbered recovery branch below only when the use must assert E.18.3 qualification, compare or publish the structure, or support a stronger downstream claim. It preserves the membership and current-result distinctions; it is not a prerequisite for understanding or correcting an ordinary route-like card.

1. Recover one selected A.22.CGUS and its four exact identity discriminators; do not create a reciprocal E.18.3 structure.
2. Name the current transformation subject or subjects, their kinds and the exact E.18 positions and bindings used by the question.
3. Classify the independently identified E.18 substrate used by selectedCGUSRef as one TFS with its valuations, one parent-relative internal SubflowRef, or one E.18.NET network of independent members and exact crossings; do not resolve the substrate ref to the selected CGUS.
4. Discriminate every continuation basis before judging a candidate. Keep an applied condition claim with its test, applicability, case inputs, and facts; keep a GuardFail as an E.18 event with its E.18/A.21 assignment facts; and use a relation-reference episteme only for an independently defined obtaining relation occurrence. For each candidate, record the dependent selected occurrences, window, outcome, and reason in a ContinuationJudgementResult, then derive the current continuation set. Carry a relation signature only when declaration-level replay needs it.
5. For each neighboringValueUseRows[] entry, recover the independently identified neighboring value through its exact kind and ref and one already-obtaining supporting relation. If the row makes a stronger claim, state in ordinary content-bearing language what the neighboring content contributes; a bare label such as test or method is not enough. A definition, constraint, predicate, test, evidence rule, or assurance rule may supply the applicable criterion, with current facts or evidence showing that the claim obtains. A Method contributes a reusable way of doing and its applicability or bounds, and a MethodDescription may state that content; any claim that its use produces, supports, evaluates, evidences, or assures a result still needs a separate applicable rule and current basis. Require an exact claim-bearing episteme, ClaimGraph, edition, or other content identity only when that identity changes the selected stronger use, and reuse an existing exact ref when available. Cite a relevant pattern only when it locates that content. A result label, return arrow, or comparison layout is not the relation.
6. Name the ordinary stop and reconsideration conditions. For a post-qualification description or demonstration, state preserved or omitted structure and add C.33 only when carrier loss affects a declared architecture use. Choose exactly one complete locator family for a one-TFS or network demonstration, or neither for a generic slice.
7. If an A.22 discriminator, CGUS locus binding or potential-continuation row, E.18 flow binding, or direct relation is missing, do not claim E.18.3 membership. If only a fact or test result for the current case is missing, keep the structure and return that candidate as unknown. If only a description-loss or stronger-use value is missing, narrow that use rather than demoting the structure.

The ordinary branch and conditional recovery sequence guide use of the pattern. They are not a local mantra, U.Method, U.MethodDescription, WorkPlan, or performed Work; completing the rows admits nothing by itself.