Obvious, it is not practical and not scalable to use one partition links
one account., we are going to use partition sharing scheme where one topic
partition will be shared by a group of accounts, which means in account
consumer, we have to check in user message level (not kafka message level)
to filter the message destination whether it belongs to the account or not.
It is too much coupling to the user application, I am wondering if there
are any measures we can use kafka message metadata to figure out the
destination. Is it possible for a kafka consumer to use kafka metadata to
multiplex the groups sharing one partition ID?