Converting callback-heavy networking to Swift Concurrency can push cancellation, isolation, and scheduling issues into production — many of these problems only appear under real load.

• Audit actor boundaries: use actor for long‑lived mutable caches and reserve @MainActor for UI‑critical code to avoid unexpected main‑thread serialization.
• Treat Task.detached as an escape hatch; prefer TaskGroup for structured parallelism so cancellation and priority are more likely to propagate predictably.
• Bridge incrementally with withCheckedThrowingContinuation shims and gate changes behind feature flags so you can roll back quickly if needed.
• Add XCTest async tests that assert cancellation behavior, and correlate test failures with OSSignposter spans and Instruments traces during pilot rollouts.

When modules mix concurrency models, prefer incremental, observable rollouts unless you can audit and instrument the full call chain end‑to‑end.

Callouts to use: TaskGroup, OSSignposter, and Instruments Time Profiler for measurement and correlation.

Are you running a migration pilot now? What signal triggered your last rollback?