Eyelink data timing

I have an Eyelink system set up to sample eye position at 1000 Hz. Looking at the MWorks event data, I have many (~5%) successive raw eye position updates with time differences of less than 500 microseconds. Inspecting the interval histogram shows a small peak around 50 us. I’ve read the timestamping article in the knowledge base, and I’m not sure what could be causing this.