I've encountered this in a couple of situations.
On a large cluster, I clicked details from the query page for an executing insert query – my browser spun and never returned – I think I may have done tried clicking a number of additional times and may have potentially exhausted the UI threads. The entire UI stopped responding to requests eventually.
On a smaller toy cluster, I ran select * from sample_08 where sleep(60000) and noticed the details link didn't respond until the query finished. I realize sleep is sorta busted, but is it possible the UI threads are waiting for something to happen in the query processing?