[BUG] Single 2-minute agentic request consumed ~80% of 5-hour usage limit

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?

A single short agentic session (~2 minutes) consumed approximately 80% of my 5-hour usage limit. The session involved Claude Code reading project files and running basic bash/search commands — no large outputs, no multi-agent parallelism, no extended thinking intentionally enabled. The usage consumed is wildly disproportionate to the actual work performed. (Claude Opus 4.8)

What Should Happen?

A brief file-reading and search session should consume a small fraction of the 5-hour limit, consistent with the actual tokens processed.

Error Messages/Logs

None — no errors were shown. The usage spike was only discovered after the session completed.

Steps to Reproduce

1. Open Claude Code with Opus model selected
2. Ask a broad codebase exploration question (e.g. "review my user management system and propose improvements")
3. Allow Claude Code to run file reads and bash/search commands for ~2 minutes
4. Check usage at claude.ai/settings/usage — observe disproportionate consumption

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.178

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Windows Terminal

Additional Information

_No response_