[BUG] Unexplained token consumption spike since early April with no change in usage pattern

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?

Starting around early April 2026, my daily token consumption increased dramatically with no change in how I use Claude Code. Same types of tasks, same repos, same session lengths, same prompts — but daily totals jumped from the ~100-400k tokens/day range I had sustained from December through late March, to repeated 2-3M tokens/day spikes.

The breakdown across models shows the increase on every model I use (Opus 4.6, Opus 4.7, Opus 4.5, Haiku 4.5, Sonnet 4.6), so it does not look tied to a single model's billing.

I have attached a screenshot of the Models tab from the usage dashboard (All range) showing the pattern clearly: flat baseline through Q1, sudden step-up starting ~Apr 10, sustained at 5-10× baseline since.

Nothing changed on my side: same machine, same workflows, same project sizes, no new heavy automation.

What Should Happen?

Token consumption should track actual usage. A 5-10× jump with no change in behavior suggests a billing or cache regression somewhere in the recent Claude Code releases.

Error Messages/Logs

No errors — silent over-billing only visible on the usage dashboard.

Steps to Reproduce

Not a reproducible single action — it is a sustained pattern visible on the usage dashboard. Happy to share anonymized session-level data if that helps the team.

Claude Model

Not sure / Multiple models

Is this a regression?

Yes, this worked in a previous version

Last Working Version

2.1.87

Claude Code Version

2.1.117

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Windows Terminal

Additional Information

<img width="547" height="457" alt="Image" src="https://github.com/user-attachments/assets/25115931-502b-4d25-b153-1a2ee65c679c" />