[BUG] Rate limits blocking multi-agent Claude Code workflows even at highest paid tier

Preflight Checklist

• [x] I have searched existing issues
• [x] This is a single bug report
• [x] I am using the latest version of Claude Code

What's Wrong?

Running multiple Claude Code instances in parallel (5-6 concurrent, not always simultaneous) on the highest paid plan tier. Frequently hitting:

API Error: Server is temporarily limiting requests (not your usage limit) · Rate limited

The error message itself distinguishes this from usage/token budget — this is infrastructure-level request rate limiting, not a monthly cap.

Frequency has been increasing. Makes multi-instance Claude Code work effectively non-viable for sustained use even at the top paid tier.

What Should Happen?

For a paying customer at the highest tier running multiple concurrent Claude Code instances:

1. Request rate limits should scale meaningfully with paid tier — top tier should accommodate parallel-agent patterns
2. Or: limits should be transparent + predictable so users can plan around them — current behavior is opaque + intermittent
3. Or: rate limits should account for parallel-agent patterns explicitly (the direction agentic coding is moving)
4. At minimum: error path should be actionable (which limit, current consumption, when it resets, what would help)

Error Messages/Logs

API Error: Server is temporarily limiting requests (not your usage limit) · Rate limited

Steps to Reproduce

1. Run 5-6 Claude Code instances concurrently on the highest paid plan tier
2. Encounter intermittent rate limit blocks at unpredictable intervals
3. Note: error explicitly states "not your usage limit" — this is request-rate limiting, not token budget

Claude Model

Opus

Is this a regression?

I don't know

Claude Code Version

2.1.142

Platform

Anthropic API

Operating System

Other Linux

Terminal/Shell

Other

Additional Information

Strategic context:

Multi-agent agentic coding is the emerging pattern for serious developer workflows. Running multiple concurrent Claude Code instances should be a strength, not a friction point. I do not experience comparable rate limit blocks on OpenAI Codex with similar multi-instance usage. This is a real competitive vulnerability as multi-agent coding scales as a market segment.

Clarification re: the public conversation about rate limits:

This report is about infrastructure request-rate limiting, NOT monthly token/usage budget caps. The two get conflated frequently. Request-rate is what blocks productive work in the moment on a paid tier; token-budget is what tier selection covers. Conflating them in user-feedback discussions misses the actual pain.

I want to raise this clearly because the conversation about rate limits often defaults to the wrong layer.