API Rate limit errors started appearing with no changes to usage patterns

Description

Starting today, Claude Code is returning API Error: Rate limit reached constantly, even for basic operations. This is new behavior — the same usage patterns (spawning subagents, running Ralph Wiggum loops with --dangerously-skip-permissions --max-turns 300) worked fine for weeks without hitting rate limits.

Environment

• Claude Code version: 2.1.80
• Model: claude-opus-4-6 (1M context)
• OS: Windows 11
• Plan: Max subscription

What changed

Nothing on my end. Same scripts, same patterns, same project. The rate limit errors started appearing mid-session and now affect:

• Subagent spawning (even 1-2 agents hit limits)
• Ralph Wiggum loops (PS1 scripts that spawn claude CLI repeatedly)
• Regular conversation when agents are running in background

Reproduction

1. Start a Claude Code session with --dangerously-skip-permissions
2. Launch 2+ background agents via the Agent tool
3. All agents immediately fail with API Error: Rate limit reached
4. Even the main conversation becomes rate limited

Impact

• Background agents that were mid-task get killed with no work saved
• PS1 loop scripts retry every 5 seconds, making the rate limit worse
• Hundreds of wasted iterations overnight (600+ iterations hitting rate limit, burning quota)
• Cannot launch parallel agents which is a core Claude Code workflow

Expected behavior

• Rate limits should match the advertised Max plan limits
• If rate limited, the CLI should back off automatically instead of returning instantly and retrying
• Spawning 2-6 subagents should not immediately exhaust burst quota — this worked fine previously

Workaround

None currently. Waiting for the limit to reset (appears to be a few minutes) then trying again with fewer parallel agents.