M2M token usage seems wrong on Auth0 platform

We have 7 APIS authenticating with M2M tokens. The token duration is 1 day, and when a token is fetched, it’s cached in the application. Since each APIs run on 2 instances, we have actually 14 servers making calls to auth0. So per month we should have around 14 * 30 = 420 tokens generated. But we have more than that (almost 1000) and since we are on the free plan it’s threatening our services to simply stop working.

I checked in the logs, and the number of calls to get a token does not correspond to what display usage on auth0 platform.

For example, on the 23rd of Octobre 2025 I have 5 calls reported, but in my logs I only have 2 calls which makes sense (1 call per service instance per day).

Thank you for your help

Actually, I made a mistake. My services are sometime authenticating to the token server to contact different service. So I generate more token that I thought, but after logging it correctly I still have 5 calls reporte instead of 4.

Calls have this body
"body": {
"client_id": "xxxxxxxxxx",
"client_secret": "[Redacted]",
"audience": "https://api.xxxx.fr",
"grant_type": "client_credentials"
}

Could it be that calls from the contacted services that check if the token is correct are counted as well ?