[BUG] Token usage depleted rapidly after 5 hour window reset with barely any 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 just had my usage limit set to 100% after the 5 hour reset.

I was in the middle of a session within visual studio code (via extension).
The session was ~30-40 minutes old.
My usage was at 40-50% within the 5 hour limit and ~35-37% of my weekly limit I think.
All these I checked roughly 1-2 minutes before the reset. then once after the reset and I noticed that my 5 hour usage was still at 37% after the reset.
Since I assumed this was a visual bug I went to the web ui, meanwhile the session kept running, doing the same things it did before. Reading some code and writing a documentation of what was present in it.
On the website I saw my usage was at 50%, while looking at it I saw each upate tick adding ~10-20% until it was 100%, it then spent 3€ in credits and was done with the task.

This is my report from this day on usage:

100% of your usage came from subagent-heavy sessions
Each subagent runs its own requests. Be deliberate about spawning them — and consider configuring a cheaper model for simpler subagents.
87% of your usage was at >150k context
Longer sessions are more expensive even when cached. /compact mid-task, /clear when switching to new tasks.
73% of your usage came from sessions active for 8+ hours
These are often background/loop sessions. Continuous usage can add up quickly so make sure it is intentional.
24% of your usage came from subagents under "general-purpose"
If this runs frequently, consider configuring its subagents with a cheaper model or tightening their prompts.

I worked in two sessions over the day, a big programming task. One I had still open from yesterday that was already big context wise. That should be the 73%.
The 27% percent left must be my fresh 40 minute session. So the usage jumping to 100% and then using credits can't be true.
The session right now has no warning for high token usage or even the little preview, so it can't have taken much.
The whole thing happened within maybe at most 5 minutes. I wouldn't even know how I could use up my entire 5 hour window in the span of 5 minutes.
The model was Sonnet, effort High.
Additionally, my weekly limit grew to 42%. Not sure how much 100%+ 5 hour limit translate to, so I'm not sure if that is realistic with the 35-37% before the reset.

Sadly there is no way I know of to figure out what exactly was supposed to have used all these tokens, so I can't really give any more detailed information.

As a comparison, I used the bigger session today, probably around 3 hours relatively frequently and was at ~35% at the end of my 5 hours. So the new session using the same model/effort at an almost empty session and causing over 100% (100% + 3€ credits) within a tiny span of time is just not possible.
The 40 minute session was also already running while the reset happened, and the usage jumped up massively right after the reset was supposed to happen.

What Should Happen?

The usage should not suddenly spike at factor of 200 of my normal usage rate.
Also, my 5 hour reset should reset my usage.

Error Messages/Logs

Steps to Reproduce

Frankly... no idea. I never had this before.
I guess for me it was starting a session ~40mins before 5 hour reset and let it work while the reset is happening.

Claude Model

Sonnet (default)

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

VSC Extension

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

VS Code integrated terminal

Additional Information

_No response_