As a fun experiment, I took four popular voice assistants – Siri, Google Assistant, Alexa, and Cortana – and put them in an infinitely looping conversation. The video below takes one iteration and plays it on loop - otherwise the video would be too long to upload!
The only times when the conversation stopped was when one of the machines wasn't able to fully pick up what another said. This can be considered an "error" on the machine's part, but is actually quite necessary to closer mimick human behavior in that humans can detect when they are in an endlessly repeating situation that leads nowhere and find ways to navigate out of it. In a way, it's like a Turing test, except the machine is the one responsible for discerning whether something is machine or human.
You can see the full, un-looping video here, where I had to cut off the fiery and passionate conversation after 10 minutes or so.