Anyone else burning through Codex tokens insanely fast today? Feel like something is wrong
I opened up a Codex session this morning for what should have been a relatively minor multi-file refactor. We're talking maybe 20–25 files max, just running a code audit review ( like i always do)
Literally after about 3 or 4 messages, I looked at my dashboard and I’d burned through over 500k tokens. My 5-hour usage limit dropped from around 70% straight down to zero, and the CLI just choked out.
It feels like the context compaction or the tool layer is acting completely broken today like it's recursively re-reading the entire file tree on every single turn instead of leveraging the cache properly. I usually use this exact workflow and barely make a dent in my allowance, but today it’s acting completely unhinged.
Are the token burn rates completely broken for anyone else right now, or did my local setup just trigger some infinite context loop?