OISST response to daily buoy temperature changes rose across 2007 but fell by one third between 2016-2019 and 2022-2025

We tested whether lower fast variance in NOAA OISST v2.1 comes only from removal of uncorrelated error. We fitted the OISST response to matched buoy temperature increments at five time intervals. At 67 stations, the median daily response ratio for 2022-2025 versus 2016-2019 was 0.666, with a spatial-group bootstrap interval of 0.618-0.785. The ratio was 0.962 for 30-day increments. The daily-to-30-day response ratio fell on a fixed set of 57 stations. Across 2007, the daily response instead rose by 16 percent at 133 stations while residual variance fell. Both signs remained on 61 common stations. Recent correlation fell at NDBC stations but rose at tropical moorings. These results show that variance loss has different covariance changes across product eras. Adding independent noise can match variance without restoring the response to measured buoy changes. This is an observational station test. Point-to-grid differences, buoy error, assimilation, changing climate states and several product changes prevent causal attribution or a global estimate of lost ocean signal. Code, sufficient statistics and verified input retrieval instructions are included.