This output can only publish using a single pre-configured routing key value; some RabbitMQ plugins highly relevant for the users of this output, e.g. rabbitmq_sharding and rabbitmq_consistent_hash_exchange which allow for workload parallelization beyond a single queue, assume that routing keys are used for message stream "sharding" (distribution).
If all routing keys are identical, all messages will end up on a single shard.
It would be very nice to have a way to tell the plugin to use random values instead of a hardcoded routing key, possibly produced by a user-provided Proc (lambda, anonymous function, etc).