[imMamdouhaboammar]

Scope

R1 governance evidence and R3 reproducibility only. No accepted ADR, runtime code, public interface, research user files, or production policy is changed.

Changes

• Add a research-owned accepted-ADR index integrity test: current canonical details and indexed IDs agree, and missing, duplicated or Proposed index entries are rejected.
• Update DEBATES with research-owned versus application-owned authority alternatives, the proposed acceptance sequence and its falsifiers.
• Record a fresh-process 1k/10k/100k EventMachine memory/time matrix for initial future events, single-event fan-out and due events in the existing R3 evaluation, with a reference to application commit 2741aaa.
• Keep the R3 resource envelope explicitly Proposed; no resource quota or streaming mode is accepted or implemented.

Verification

• python3 -m pytest -q: 63 passed (baseline 58).
• git diff --cached --check: clean.
• Application benchmark smoke tests: 4 passed; app full Python suite: 586 passed and 58 subtests.
• The three 100k-event workloads had approximately 88-89 MB process peak RSS on one CPython 3.12 arm64 macOS run each. Not a memory guarantee or 100k-agent benchmark.

Review gates

• Owner must choose and accept canonical ADR authority in a new immutable decision; this proposal does not make that choice.
• Research-only drift is not blocked atomically by application CI. Application Draft PR Research: restore ADR index integrity and document resource bounds #7 proposes an eventual nightly parity probe: Recovery: align architecture governance and fail closed in Web Studio funnelscript#7
• R3 opt-in limits, typed failure semantics, streaming alternatives and later legacy convergence remain gated on accepted contracts.
• The existing uncommitted changes in the main research checkout were not accessed or modified.