Currently, none of the TCP-based stacks have UFC in them. This is bad, as async messages can overwhelm the receiver. Point in case is cross-site replication: the bridge using TCP replicates updates to the other site masters quickly, but the updates are only passed to the thread pool for processing, exhausting memory over time.

Although TCP does perform flow control, the application code at the receiver may delay delivery and thus the sender would block if UFC was present.

However, the absence of UFC leads to thread pool (and queue) overflow at the receiver(s) if messages are constantly sent at a rate greater that when the receiver(s) can process.