I would imagine this will become more and more common with reasoning LLMs that take a while to return their responses and output interim chain of thoughts.