fix(openclaw): quiet duplicate typed hook timeouts

[Jurij89]

Summary

• Preserve per-API typed hook installs so before_prompt_build and agent_end remain reachable across OpenClaw multi-phase registries.
• Add aggregate-aware typed hook commitment so sibling surfaces that see the same typed event fire elsewhere commit quietly instead of emitting duplicate commit-by-timeout warnings.
• Keep frequent hooks frequent, keep rare-fire treatment limited to rare hooks, and keep typed install failures warning-level.

Related

• Fixes adapter-openclaw: typed hook commit-by-timeout warns fire 2× on multi-phase init even when hooks are active #338
• Related to Fix OpenClaw hook log noise #336
• Related to fix(openclaw): align DKG home and state-dir resolution #337

Files changed

File	What
packages/adapter-openclaw/src/DkgNodePlugin.ts	Records typed hook fire generations with return-preserving wrappers and passes per-install peer-fire observation callbacks.
packages/adapter-openclaw/src/HookSurface.ts	Adds committed-by-peer-fire timeout commitment for typed hooks observed on sibling retained surfaces.
packages/adapter-openclaw/test/plugin.test.ts	Adds regression coverage for multi-phase duplicate timeout suppression, typed install failure warnings, setup-runtime retry coverage, and typed rare-fire assertions.

Test plan

• pnpm install --frozen-lockfile
• pnpm --filter @origintrail-official/dkg-core build
• pnpm --filter @origintrail-official/dkg-adapter-openclaw test -- test/plugin.test.ts -t T338|T6|T31|T59|rare-fire
• pnpm --filter @origintrail-official/dkg-adapter-openclaw exec vitest run test/plugin.test.ts test/HookSurface.test.ts
• pnpm --filter @origintrail-official/dkg-adapter-openclaw run build
• pnpm --filter @origintrail-official/dkg-adapter-openclaw exec vitest run test/before-prompt-build-hook.test.ts test/ChatTurnWriter.test.ts
• pnpm build:runtime
• git diff --check

[github-actions]

Codex review completed — no issues found.

[Jurij89]

CI/review sweep after the PR Driver wait:

• Codex PR Review and Knip passed; there are no unresolved review threads.
• The repo-wide CI workflow is red, but the failures I inspected are outside this PR's adapter-openclaw change set. Examples include network-sim tests explicitly named as current parity/seeded-RNG failures, agent endorsement/signature A-7 tests marked as currently failing in the suite, CLI daemon/auth/keystore failures, and storage/publisher shards. The Build packages job passed, and the local OpenClaw-focused verification listed in the PR body passed.

I'm leaving the failures classified as unrelated to #338 rather than broadening this PR into unrelated package fixes.