Way to detect invalid device token in FCM HTTP v1 API - firebase-cloud-messaging

Question

we are using FCM HTTP v1 API to send push notification.
When, our users register their device tokens, our push server just stores it to database without verification because FCM does not provide token verification APIs.
The only time the push server could detect whether the device token is valid or not is when sending push notification via FCM.
According to FCM registration token management, if the device token is invalid the FCM server responds with UNREGISTERED or INVALID_ARGUMENT.
However, the INVALID_ARGUMENT code can be returned when using invalid payload as well. So we can not distinguish errors between invalid device token and invalid payload.
The following is actual response from the FCM server in case of using invalid device token and invalid payload respectively.
{
"error": {
"code": 400,
"message": "The registration token is not a valid FCM registration token",
"status": "INVALID_ARGUMENT",
"details": [
{
"#type": "type.googleapis.com/google.firebase.fcm.v1.FcmError",
"errorCode": "INVALID_ARGUMENT"
}
]
}
}
HTTP/1.1 400 Bad Request
{
"error": {
"code": 400,
"message": "Invalid JSON payload received. Unknown name \"priority\" at 'message': Cannot find field.",
"status": "INVALID_ARGUMENT",
"details": [
{
"#type": "type.googleapis.com/google.rpc.BadRequest",
"fieldViolations": [
{
"field": "message",
"description": "Invalid JSON payload received. Unknown name \"priority\" at 'message': Cannot find field."
}
]
}
]
}
}

Answer 1

When the token has a proper format but its user is no longer active, you will receive an UNREGISTERED response indicating that you should delete that token from your backend.
However, if you send a token that does not have a valid format (it has been modified in your backend or truncated, for instance) you will receive an INVALID_ARGUMENT. If this is the case you probably want to check your code because you may be modifying the token somewhere (in App or backend). Probably you have to manually detect and delete those tokens.
Once this is solved, you do not have to worry anymore about the INVALID_ARGUMENT response to delete the tokens, except if you make further changes that corrupt them again.