I ran the range_compare script using the 10-12 minute low range stretch from the most recent lock and some time before it of the same lock where the range was better, the construction crew by staging was also seen moving large equipment around this time on the cameras.

To produce this figure I ran "python3 /ligo/gitcommon/ops_tools/rangeComparison/range_compare.py --span 600 1415725641 1415728983", I only did 600 seconds of data. The range drop looks to be from low frequency noise, mostly below ~300Hz, the most obvious peak difference looks to be at 46/47 (mostly on sensmon), 36 (mostly on sensmon), and 60Hz (both sensmon and DARM) for the low range stretch. The lines are hard to see, but if you open the pdf in LibreOffice (linux default pdf viewer) and right-click -> Arrange -> Behind Object, then you can hover your cursor over the curves and it will be much easier to see the differences.