Next findinds:
I am writing to share new data and findings. Here is the second reason why I am absolutely certain that a “phantom” resource was stuck on your side, and why this case needs to be reviewed by your engineering team.
Currently, my Gemini API has been completely Disabled for three days (since March 18). Interestingly, day by day, those massive cache storage charges are now literally disappearing from my billing graphs.
If you look at the pre-disable graphs for March 15-17, the system was eating about $36 per day. However, after I force-disabled the API, the billing system started retroactively “recalculating” the cache costs back to the actual, correct data.
**Billing Support mentioned a 32-hour propagation delay, but for some reason, this synchronization lag took about 3 days to correct itself. And possibly it’s because I Disabled API. It started cleaning ONLY after I DISABLED API.
**
Furthermore, in a live production environment, I wouldn’t be able to simply disable the entire API to force the billing system to sync. So please check and fix this issue.
My current graph looks like (orange).
I really hope your engineers can fix this frightening delay issue for the Gemini Cache API.