80 meter noise, part 2

I took a look at the audio capture of the 80M noise, and here's when it's happening. Entries are start time (local PST), duration, comments.

• 0900-1400 - quiet, no noise
  
• 1400 (2m 22sec) - first time heard today, stops abruptly
  
• 1522 (3 min 15 sec) - fades out slowly
• 1535 (14 min 5 sec) - stops abruptly
• 1556 (7 hr, 35 min, 7 sec) - longest on period. Very strong, S9. Ends abruptly at 2331 local. Here's the very beginning of this.
  
• 2331 - 0829 - this is a long period where the noise appears at a reduced amplitude (probably S1 to S3) for a period (30 - 90 seconds) and disappears for several minutes. The pattern tends to be regular, e.g. at about 0300 local time, the pulses come about every 18 seconds, with a duty cycle of about 50%. Here's a sample.

For reference, local sunrise is at 0700 and local sunset is at 1652 today. This pattern is rather odd. The long "on" period from about 4pm to 11:30pm local time really sounds like something humans would do when it gets dark, but I'm a bit stumped by the repeating patterns all night, which finally seem to disappear about 8:30am (which is when you might expect people to be out of a house).