#19551: fix(reply): make overflow fallback deterministic for empty recovery payloads

## Summary - Problem: overflow recovery had edge paths that could finalize with no renderable payload (undefined/empty outcome) when `meta.error` was `context_overflow`. - Why it matters: this creates silent/non-actionable behavior in a high-friction reliability path. - What changed: - Added deterministic overflow fallback when embedded overflow meta-error returns zero payloads. - Added deterministic overflow fallback when payloads normalize/filter to empty in `runReplyAgent`. - Added two regressions covering both paths. - What did NOT change (scope boundary): no config/schema changes, no provider selection logic changes, no broad compaction refactor. ## Change Type (select all) - [x] Bug fix - [ ] Feature - [ ] Refactor - [ ] Docs - [ ] Security hardening - [ ] Chore/infra ## Scope (select all touched areas) - [x] Gateway / orchestration - [ ] Skills / tool execution - [ ] Auth / tokens - [ ] Memory / storage - [ ] Integrations - [ ] API / contracts - [ ] UI / DX - [ ] CI/CD / infra ## Linked Issue/PR - Closes #19550 - Related #1187 - Related #18095 - Related #18997 - Related #15749 - Related #18663 - Related #18719 ## User-visible / Behavior Changes - Overflow edge paths now return a deterministic fallback response instead of potentially finalizing with an empty/undefined payload. ## Security Impact (required) - New permissions/capabilities? (`No`) - Secrets/tokens handling changed? (`No`) - New/changed network calls? (`No`) - Command/tool execution surface changed? (`No`) - Data access scope changed? (`No`) - If any `Yes`, explain risk + mitigation: `N/A` ## Repro + Verification ### Environment - OS: Ubuntu 25.10 - Runtime/container: local source checkout - Model/provider: N/A (runner behavior) - Integration/channel (if any): reply-agent execution path - Relevant config (redacted): defaults ### Steps 1. Return `meta.error.kind = context_overflow` with `payloads: []` and unavailable session reset context. 2. Return overflow error payload that normalizes/filters to empty. 3. Observe finalization behavior. ### Expected - Deterministic fallback payload is returned in both edge paths. ### Actual - Before fix: empty/undefined-style finalization possible. - After fix: explicit overflow fallback response returned. ## Evidence Attach at least one: - [x] Failing test/log before + passing after - [x] Trace/log snippets - [ ] Screenshot/recording - [ ] Perf numbers (if relevant) Regression tests added: - `returns a deterministic overflow fallback when error meta has no payloads` - `returns a deterministic overflow fallback when error payload normalizes to empty` ## Human Verification (required) What you personally verified (not just CI), and how: - Verified scenarios: - `pnpm vitest run src/auto-reply/reply/agent-runner.runreplyagent.test.ts` passed (25 tests). - `pnpm test` passed. - `pnpm build` passed. - Edge cases checked: - existing compaction-reset and role-ordering tests still pass. - What you did **not** verify: - exhaustive channel-by-channel delivery matrix. Note: `pnpm check` reports pre-existing repository-wide formatting issues (`Format issues found in above 1121 files`) unrelated to this PR's touched files. ## Compatibility / Migration - Backward compatible? (`Yes`) - Config/env changes? (`No`) - Migration needed? (`No`) - If yes, exact upgrade steps: `N/A` ## Failure Recovery (if this breaks) - How to disable/revert this change quickly: - Revert this PR commit. - Files/config to restore: - `src/auto-reply/reply/agent-runner-execution.ts` - `src/auto-reply/reply/agent-runner.ts` - `src/auto-reply/reply/agent-runner.runreplyagent.test.ts` - Known bad symptoms reviewers should watch for: - reappearance of empty/undefined finalization behavior on overflow edge paths. ## Risks and Mitigations - Risk: fallback could trigger in cases where empty payload was intentional. - Mitigation: fallback is restricted to explicit overflow/compaction error kinds. - Risk: overlap with adjacent overflow/compaction work. - Mitigation: keep patch narrowly scoped and link related issues/PRs. ## AI assistance - AI-assisted: Yes (author verified code paths, tests, and scope).