OpenCode native continuation takeover — current-head proof
Source: https://github.com/pingdotgg/t3code/pull/19388
Bottom head: b2d5051b10d508381ca44d3e4ad86ffd4941432c
Top head: 5c93a777fa8dc084332ed5c46f4259b3303747eb

Focused baseline: 154 tests in OpenCode2AdapterV2.test.ts and RunExecutionService.test.ts.
Retained coverage: 156 tests in those same files. One new two-variant service case, two strengthened adapter cases.
Before production fix: 152 pass, 4 assertion failures; queued child running/output/completed missing in both ownership variants; pending snapshots absent in two adapter scenarios.
After production fix: 156 pass, 0 fail.
Bottom-only adapter suite: 100 pass.
Parent integrated focused run on top head: 4 pass, 152 filtered, exit 0.
Relevant provider-core/provider-opencode/server typechecks and targeted lint/format pass.
Independent full-diff reviews: Claude Opus 5.5 and Codex GPT-6 Sol, one clean round.
Original nested Stop stale ancestor concern fixed and regression verified.

Native OpenCode 2.0.28 isolated server started and API model catalog worked.
A zero-cost native model created a real child. Its general agent denied nested subagents, so the queued grandchild continuation was NOT reproduced.
No claim of end-to-end native continuation proof. The isolated web app paired and loaded its generated project/thread; browser host font rendering prevented useful visual evidence. No UI source change.
macOS, Windows, iOS, Android, Electron desktop unverified.
Human provider verification and approval remain required. No merge.