COOL IDEA: deterministic multi-writer scenario corpus

[flyingrobots]

Idea

Build a source-controlled corpus of deterministic multi-writer graph scenarios that can be reused across reducer, materialization, optic, sync, and documentation-example tests.

This is the constructive version of #681: instead of ad hoc random fixture generation inside integration tests, name the causal scenarios we care about and make them replayable.

Shape

Each scenario should describe:

• writers and Lamport timelines
• patch operations and expected dots
• intended causal relationships
• expected materialized state
• expected index/read-basis observations where relevant
• regression names explaining what invariant the scenario protects

Why it is cool

A corpus like this gives us reusable behavioral truth:

• deterministic CI failures
• cheap cross-path equivalence tests
• fixture-backed docs examples
• stable coverage paydown targets for Coverage ratchet baseline dropped during v17 release preflight #261
• future optic/strand/merge scenarios that can be replayed without inventing a new harness each time

Suggested first slice

Start with 5 to 8 compact scenarios:

• independent writer convergence
• concurrent property overwrite
• node remove with observed dots
• edge add/remove race
• checkpoint then post-checkpoint continuation
• divergent writer branch with common basis

Then wire one consumer, likely JoinReducer, before expanding to materialization or optic readers.