Define Ultra degradation and usage guardrails

[iam-brain]

Question

What deterministic behavior and user feedback should apply when Ultra cannot provide normal proactive delegation or could create excessive work?

Resolve with the maintainer:

• missing or disabled task/subagent tooling;
• collaborationProfileEnabled or orchestratorSubagentsEnabled disabled;
• unsupported runtime/agent contexts;
• child-agent spawn failure, cancellation, or partial completion;
• nested delegation and recursive fan-out;
• low and high concurrency settings;
• rapid quota consumption, account failover, and weekly/5h exhaustion;
• whether the plugin hides Ultra, rejects the turn, warns and falls back to Max, or continues in another explicit state;
• where warnings appear, how often they repeat, and what debug metadata records;
• whether a default concurrency recommendation or enforceable cap belongs in plugin configuration.

The resolution must define state-by-state behavior that is consistent with the user-visible Ultra contract and honest about best-effort adaptation limits.