Analyze persistent Workshop object lifecycle and fan-out

[Teakowa]

Goal

Make Wright surface persistent Workshop object lifetime and fan-out risks as structured analysis so creators can identify visual/object allocations that can accumulate, multiply across players, or lack an evident cleanup path.

Context

Wright already has a small evidence-classified stability rule set for loops, repeated evaluation, and expensive predicates, but it does not currently model persistent Workshop objects such as HUD text, in-world text, or effects as resources with creation scope, visibility/fan-out, and cleanup responsibilities.

This leaves an important Workshop-specific stability class to manual review. Community Workshop guidance repeatedly treats persistent visual objects as a distinct risk because one creation site may multiply by player count, remain live across state transitions, or be recreated without an evident destruction path. The useful product capability is not a universal "too many effects" threshold; it is structured static evidence about what is created, where, how broadly it can fan out, and whether its lifetime is evident.

Scope

• Analyze canonical Workshop object-creation operations whose products persist until explicitly destroyed or otherwise have a meaningful lifetime contract.
• Report creation sites with source provenance and the rule/event scope that can create them.
• Expose statically knowable fan-out dimensions such as global vs per-player execution and broad visibility/target sets where canonical WIR makes them identifiable.
• Distinguish exact structural facts from heuristic lifecycle risks.
• Identify only defensible missing-lifetime cases, such as an object whose created identity is not evidently retained or whose relevant scope has no observable cleanup path; ambiguous aliasing/data-flow cases must remain explicit limitations rather than correctness claims.
• Keep findings useful for both raw Workshop and provider-backed source languages through Wright's canonical Workshop analysis boundary.
• Validate the feature against representative real-project object-lifecycle patterns in addition to focused positive/negative fixtures.

Non-goals

• Defining a universal safe number of HUD texts, effects, or other objects.
• Claiming precise runtime/server load from static object counts.
• Treating every globally visible or per-player object as a defect.
• Requiring full interprocedural lifetime proof before providing useful structural evidence.
• Automatically rewriting object creation, visibility, or cleanup code.
• Moving Workshop object semantics into Wright; canonical identities and lifetime facts remain owned by workshop-rs.
• Adding source-language-specific OPY/DEL lifecycle logic.

Acceptance criteria

• Wright can report supported persistent-object creation sites with stable machine-readable identity, source provenance, and event/execution scope.
• The analysis exposes statically knowable fan-out information without presenting heuristic multiplicity as measured runtime load.
• At least one stable finding/metric covers an evidenced lifecycle risk that the current rule set cannot detect, with explicit evidence classification and known limitations.
• Safe examples with an evident bounded lifetime/cleanup path are distinguishishable from corresponding risky examples.
• Per-player and broad-visibility cases are represented without assuming a fixed server population unless that population is part of explicit input/project evidence.
• Each shipped lint finding has a stable Rule ID, rationale, limitations, and positive/negative fixtures; purely descriptive metrics need not be promoted to lint findings.
• Representative real-project evidence demonstrates that the result is useful beyond synthetic fixtures.
• Raw Workshop analysis remains regression-green, and provider-backed consumers can reuse the same canonical analysis without duplicating source-language semantics.
• Independent implementation ablation removes the new lifecycle/fan-out logic and causes its dedicated regression evidence to fail again.

Dependencies / ownership

• workshop-rs owns canonical Workshop action/value identities and exact object semantics.
• Wright owns static analysis, lint classification, reporting, and cross-language product integration.
• Provider provenance integration remains governed by Converge source-language workflows on an explicit provider boundary #243/Preserve Wright lint and analyze semantics across the OPY provider boundary #246; this feature must not invent source mappings when canonical evidence cannot be mapped back to authored source.