It happened a second time on a different camera. Not the crosswalk camera. A privately-operated camera I have been accessing via a neighbor who does not know what I use the feed for. The camera is pointed at the street from the second floor, covering approximately sixty meters of the block. The gap in this feed is at a different time on a different day, but the duration is the same: three minutes and seven seconds. Not three minutes exactly. Three minutes and seven seconds, identical to the gap in the crosswalk camera footage to the second.

The probability of two independent gaps in two independent camera feeds being the same duration to the second by coincidence is not a calculation I am interested in performing because the result would be a number I would dismiss anyway. The implication is that the gaps are not camera failures or recording artifacts. They are interventions of a fixed duration. Three minutes and seven seconds is how long the intervention takes. Whatever the intervention is, it runs on a timer.

I have documented both gaps in the notebook. The gap timestamps do not overlap. The cameras are not on the same network. The recording devices are different brands, different firmware, different storage methods. The only thing they have in common is the duration of their absence and the fact that I have access to both.