[Roadmap] Portfolio strategies and scheduled rebalancing

Problem

Strategy API V2 can work with multiple instruments, but large dynamic portfolios need a documented lifecycle for universe changes, scheduled selection, target generation, execution, and restart-safe state.

Desired outcome

Make portfolio strategies reproducible and operationally safe across backtest, paper, and live execution without creating a second strategy engine.

Initial scope

• Define the lifecycle for selection, rebalance, target generation, and reconciliation.
• Support scheduled refresh using saved point-in-time universes.
• Preserve deterministic ordering and idempotency across retries and restarts.
• Define provider, broker, and runtime limits for large universes.
• Add portfolio-level diagnostics without hiding per-instrument orders.

Acceptance criteria

• An RFC documents lifecycle callbacks, state ownership, and failure behavior.
• Backtest and live runtimes share the same target-generation contract.
• Restart, duplicate schedule delivery, and partial execution scenarios are tested.
• Scale limits and rate-limit behavior are documented and observable.
• Existing single-instrument strategies remain compatible.

Out of scope

• Claiming a certified capacity for 100 pairs before load and broker testing.
• Atomic multi-leg execution; that is tracked separately.

Contributing

This is an epic. Comment with one scoped lifecycle, test, documentation, or observability task and a short implementation plan. A maintainer will confirm scope before assignment.