Archive the flat eager candidate, its direct comparison artifacts, the two deterministic round-robin runs, and the study summaries that explain the outcome. This experiment asks a concrete question: can a flatter eager point-based event layout help without changing parser behavior or paying the cost of lazy getters. The answer from both artifact paths is no. The refreshed direct comparison lands at about -1.49 percent target timing with about -3.97 percent retained memory, and the round-robin rerun only shifts timing to about +0.95 percent with 2 of 9 significant target wins. Both still miss the +5 percent acceptance bar, and both keep memory moving in the wrong direction. The reason to keep all of this in one commit is methodological. The candidate snapshot, the recorded schedules, and the study summary belong together because they show both the implementation that was tested and the evidence used to reject it. The round-robin runs also make the evaluation easier to defend by reducing simple ordering bias and by preserving the exact commands and report paths that were used. Verification: this commit records measured results and summary docs; the study-local tooling used to produce them was type-checked with `mise x deno@latest -- deno check experiments/event-shape-study/tools/*.ts`