Usage limits hit within seconds on a single prompt

Bug Description
A few days ago I was working with Claude, decided to try Opus for a couple more difficult prompts, submitted them and got great answers, but on my third prompt I hit my usage limit. I figured it was just Opus eating through my usage faster, so after waiting for the reset, I switched the model to Sonnet and re-submitted the prompt. Only to hit the usage limit before it was able to respond. I thought that maybe the sub-agents it was spinning up were still using Opus, so after waiting again, I reset to an earlier point in the conversation (Esc+Esc) (I didn't want to lose all my context) set the model to Sonnet, submitted my prompt, and hit the usage limit _again_ before I got a response. I thought maybe there was something strange going on with that session, so I grabbed the full conversation, including the previous compaction (about 80k tokens total), pasted those into a markdown file, started a brand new Sonnet session, and asked it to read the markdown file. Which it did fine. Told it to ONLY use Haiku for sub-agents. But then I submitted my prompt again and _again_ I hit my usage limit before it could respond. Maybe 5 seconds of operation? The primary session agent decided to spin up 7 sub-agents to do research, which I know is a lot of tokens, but I've done similar before without hitting the usage limit. I've seen it use maybe 30% of my limit for that type of workload.

At this point it's been 4 days trying to submit this single prompt, and I'm running out of ideas to try. I had been thinking about bumping up a usage tier, but if I can't figure out why a single prompt is churning through my limits so quickly, or if there's a bug in your system that's miscounting usage, I'm definitely not going to be spending more money on this. And I swear, the behavior changed this weekend. I've used claude extensively over the last few months at home and at work, and I've never seen it churn through my limits like this.

Unfortunately, due to an NDA, I can't submit this in either of the transcripts I've tried. But hopefully my description can help you, help me, track down what's going on and why Claude Code has suddenly become unusable.

Environment Info

• Platform: darwin
• Terminal: zed
• Version: 2.1.62
• Feedback ID: 11f893d4-a785-4a8f-805a-5401bd105359

Errors

[{"error":"AbortError: The operation was aborted.\n    at unknown\n    at abortChildProcess (node:child_process:935:42)\n    at onAbortListener2 (node:child_process:35:24)","timestamp":"2026-02-27T03:44:29.104Z"},{"error":"RipgrepTimeoutError: Ripgrep search timed out after 20 seconds. The search may have matched files but did not complete in time. Try searching a more specific path or pattern.\n    at $ (/$bunfs/root/claude:116:189)\n    at <anonymous> (/$bunfs/root/claude:116:403)\n    at exitHandler (node:child_process:109:27)\n    at errorHandler (node:child_process:118:16)\n    at emitError (node:events:43:23)\n    at abortChildProcess (node:child_process:935:17)\n    at onAbortListener2 (node:child_process:35:24)","timestamp":"2026-02-27T03:44:29.104Z"},{"error":"Error: 429 {\"type\":\"error\",\"error\":{\"type\":\"rate_limit_error\",\"message\":\"This request would exceed your account's rate limit. Please try again later.\"},\"request_id\":\"req_011CYXtHbdRjLqtxmT8XNai2\"}\n    at generate (/$bunfs/root/claude:268:28525)\n    at makeRequest (/$bunfs/root/claude:307:5435)\n    at processTicksAndRejections (native:7:39)","timestamp":"2026-02-27T03:44:34.507Z"}]