[BUG] Rate limit percentage appears incorrect after reset - shows 17% with minimal usage

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?

I'm on the Pro plan. After hitting my hourly rate limit earlier today, I waited for the reset at 5PM Pacific. Since the reset, I've only sent 2-3 short text messages (questions about rate limits), yet the CLI shows 17% usage and the web UI shows 18%. Also concerning is that after my sessions yesterday and today my weekly consumption is already at 37% and climbing fast since the 5PM reset. For similar work last week, the usage was not climbing this quickly.

Timeline

1. I was working in Claude Code (creating python notebook files, running bash commands, running some general queries)
2. Hit rate limit around 330PM, paused work
3. Waited for reset at 5PM Pacific
4. After 5PM, sent only 2-3 short text messages asking about rate limits
5. /status shows 17% used since 5PM reset
6. Web UI (claude.ai) shows 18% usage

What Should Happen?

After rate limit reset, 2-3 short text messages should consume <1% of rate limit, not 17-18%.

Error Messages/Logs

Ran some commands to locate logs, settings and project info. There are no logs or project information returned. Here is my settings file as recommended by Claude:
{
  "permissions": {
    "defaultMode": "default"
  },
  "model": "opus",
  "alwaysThinkingEnabled": true
}

Steps to Reproduce

(creating python notebook files, running bash commands, running some general queries)

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

The rate of consumption is faster than I recall from last week.

Claude Code Version

2.1.5 (Claude Code)

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

1. Do tool calls (bash commands, file writes) from before the reset carry over somehow?
2. Is the percentage calculation based on tokens, API calls, or something else?
3. Is there a delay in the reset taking effect?

<img width="732" height="433" alt="Image" src="https://github.com/user-attachments/assets/b9dfcf23-d460-4159-b29d-e31db7629541" />

I do some similar reports of folks speaking about their usage spiking but this is dependent on workload and I couldn't map that.