📖About

INI-218 added rate-limit parking in Elrond, but exec 560 on INI-209 still moved to needs_attention because SP-01 and Developer logged the usage-limit response as generic parse/codegen failures. Preserve the reset signal end-to-end so dispatcher sees it, parks the queue row with paused_until, and current execution stays parked until auto-resume.

⚖️Decisions

• If Claude says the limit was hit with a reset time, the initiative stays parked in current execution until that reset window elapses. — This is a predictable capacity event, not an escalation-worthy failure. (Ninad)
• The fix belongs in signal propagation, not in a broader change to blocked vs paused semantics. — INI-218 already established the right dispatcher behavior; the regression is that upstream droids are hiding the rate-limit signal. (Orion)
• requester rewritten ninad -> orion per CAROL-INI-744: orion is the only human-CLI requester — Backfill of historical rows after INI744 added API-level refusal of requester=ninad. Orion is Ninads CLI agent; all human-originated initiatives are filed with requester=orion. (orion)

✅Success criteria

• Unit test: SP-01 returns a rate_limit-shaped error payload with reset_at when raw Claude result is a usage-limit message. (must_have)
• Unit test: developer/codegen path records a rate-limit-shaped signal rather than plain No executable code generated when raw Claude result is a usage-limit message. (must_have)
• Integration-style test: dispatcher failure sweep pauses a dispatch_queue row with paused_until set when droid_runs contain a rate-limit signal. (must_have)
• Replay on INI-209-style data shows the detector would pause instead of writing needs_attention. (must_have)