[BUG] Massive spike in messages and abnormal session/cache behavior on February 1st

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?

The clearest indication of a bug is the massive spike in activity on February 1st. During normal operation between January 27th and 31st, the number of messages per day ranged between –5 and 791–. On February 1st, this suddenly exploded to –3,339 messages–. That is more than all previous days combined. In total, –4,563 messages– were recorded across the entire period, meaning that over –70%– of all messages originated from this single day.

The session behavior is equally suspicious. Although only –15 sessions– were recorded on February 1st, the number of messages per session was extremely high. There is a longest session containing –1,048 messages–. A single session exceeding –1,000 messages– is highly unusual and strongly suggests an infinite loop. It is very likely that an automated process, script, or agent repeatedly sent the same or very similar requests without terminating correctly.

Additionally, the cache usage is extraordinarily abnormal. The cacheReadInputTokens values are astronomically high compared to actual token usage. For the Opus model, more than –56 million– cache-read tokens were recorded, and for the Sonnet model, more than –5 million–. This indicates that extremely large prompt contexts were repeatedly retrieved from cache. This behavior is another strong signal of a failed retry mechanism, where the same large context is continuously reused within an error loop.

The temporal distribution further supports this conclusion. Activity is heavily concentrated in specific hours, such as –96 counts– in hour –5– and –29 counts– in hour –11–. This pattern is far more consistent with automated machine activity than with human usage, which would normally be more evenly distributed throughout the day.

Given that I have already seen multiple issues complaining about weekly usage limits, I conclude that this is either a server-side bug or a deliberate decision by management. For months, experimentation has been happening at the expense of users: terminal UI bugs, deliberate restrictions on open-source usage despite a Claude Max subscription. When the CTO proudly showcases his workflow publicly, I now have a clearer idea of who might be consuming my tokens in the process. At this point, I’m out.

What Should Happen?

Nothing to preview

Error Messages/Logs

"version": 2,
  "lastComputedDate": "2026-02-01",
  "dailyActivity": [
    { "date": "2026-01-27", "messageCount": 5, "sessionCount": 1, "toolCallCount": 0 },
    { "date": "2026-01-28", "messageCount": 17, "sessionCount": 3, "toolCallCount": 1 },
    { "date": "2026-01-29", "messageCount": 791, "sessionCount": 107, "toolCallCount": 75 },
    { "date": "2026-01-30", "messageCount": 180, "sessionCount": 31, "toolCallCount": 21 },
    { "date": "2026-01-31", "messageCount": 231, "sessionCount": 52, "toolCallCount": 11 },
    { "date": "2026-02-01", "messageCount": 3339, "sessionCount": 15, "toolCallCount": 412 }
  ],
  "dailyModelTokens": [
    { "date": "2026-01-28", "tokensByModel": { "claude-opus-4-5-20251101": 69 } },
    { "date": "2026-01-29", "tokensByModel": { "claude-opus-4-5-20251101": 27719, "claude-sonnet-4-5-20250929": 33 } },
    { "date": "2026-01-30", "tokensByModel": { "claude-sonnet-4-5-20250929": 1025, "claude-opus-4-5-20251101": 137 } },
    { "date": "2026-01-31", "tokensByModel": { "claude-sonnet-4-5-20250929": 3677, "claude-opus-4-5-20251101": 46 } },
    { "date": "2026-02-01", "tokensByModel": { "claude-opus-4-5-20251101": 18804 } }
  ],
  "modelUsage": {
    "claude-sonnet-4-5-20250929": {
      "inputTokens": 1960,
      "outputTokens": 2775,
      "cacheReadInputTokens": 5418631,
      "cacheCreationInputTokens": 6114880,
      "webSearchRequests": 0,
      "costUSD": 0
    },
    "claude-opus-4-5-20251101": {
      "inputTokens": 42806,
      "outputTokens": 3969,
      "cacheReadInputTokens": 56261599,
      "cacheCreationInputTokens": 11890699,
      "webSearchRequests": 0,
      "costUSD": 0
    }
  },
  "totalSessions": 209,
  "totalMessages": 4563,
  "longestSession": {
    "durationMs": 5626912,
    "messageCount": 1048
  }
}

Steps to Reproduce

Nothing to preview

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.29 (Claude Code)

Platform

Other

Operating System

Other Linux

Terminal/Shell

Other

Additional Information

<img width="1921" height="925" alt="Image" src="https://github.com/user-attachments/assets/4acb3451-2063-42d4-b5be-1790e1b3404d" />