[BUG] "API Error: Rate limit reached" repeatedly despite having significant quota remaining.

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Bug Description

Claude Code is hitting "API Error: Rate limit reached" repeatedly despite having significant quota remaining. The usage dashboard confirms low utilization. This is happening on a Saturday morning (off-peak) with freshly renewed credits.

Environment

• Plan: Max (with €150 extra usage limit, €57.22/€150 used)
• Date/Time: Saturday, March 28, 2026, ~07:00 CET (06:00 UTC)
• Session usage: 15% (resets in 3hr 3min)
• Weekly usage: 8% (resets Fri 6:00 AM)
• Sonnet usage: 0%
• Claude Code command: claude --resume 372d208e-f3fb-4dfe-8eb6-588e7de841c0 --dangerously-skip-permissions
• Token count at time of errors: ~609,855 tokens
• Model: Opus 4.6

What Happened

1. Asked Claude Code to trace case creation logic in a Python backend (trust-relay-workflow/backend/app/api/case_crud.py)
2. Claude Code started reading files and running bash grep commands — nothing intensive, no parallel agents, no large generations
3. After ~34 seconds of work, hit first API Error: Rate limit reached
4. Every subsequent action (even just Read 160 lines or Read 100 lines) immediately triggers the same error
5. Typing continue results in another rate limit error within seconds
6. This cycle repeated 5+ times in a row — completely unable to work

What Should Happen

With 15% session usage and 8% weekly usage on a Max plan during off-peak weekend hours, basic file reading operations should not trigger rate limits.

Evidence

• Dashboard screenshot showing 15% session / 8% weekly / 0% Sonnet: [attached]
• Terminal screenshot showing repeated rate limit errors on simple read operations: [attached]

Additional Context

• This is on a Saturday — outside Anthropic's stated peak-hour throttling window (weekdays 5AM–11AM PT)
• Credits just renewed
• Same workloads ran fine previously on this plan
• Aware of the recent prompt caching bug that caused similar phantom usage drain — this looks like a recurrence
• The mismatch between dashboard (15%) and actual enforcement (hard blocked) suggests the rate limiter and the usage meter are out of sync

Impact

Completely unable to work. Paying for Max plan and getting zero value. This is a production workflow blocker.

What Should Happen?

I should be able to work, critical projects are on hold!

Error Messages/Logs

Steps to Reproduce

just resumed my session, "continue" and had api rate limit almost instantly a few times

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.86

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_