Since publishing "FCM HTTP v1" Google improved security system of their FCM services. Now it requires to use short-time access tokens of OAuth 2.0 service, which uses private key based authentication and specific Google software libraries.
Each terminal vendor has it's own application with it's unique Application ID registered and Google Firebase. Terminal may have several applications with different IDs. It should not be a good idea to hardcode Application IDs inside BM's code as well as vendor's private key to pass Google Authentication.
Due to real-time nature of the service all messages should be sent by BM directly. This should solve issues of inaccessability of service and power of vendor's site.
When terminal goes to the idle mode it may pass IDLE message to BM contains info about Vendor's service which will provide information to pass messages to FCM such as application identifiers, access token and expiration time.
On this way terminal vendor delegates message sending feature to BM without disclosing information of its private key.
BM / Registry will store these data and assosiate it with particular terminal / connection context and use it to wake application up.
Terminal may add a verification string to URL of token service to check that request of token initiated by vendor's application, it could be something like CRC32 hash of BM's IP salted with secret known only to terminal and token service.
BM will call token service only in following cases: