I’m seeing the same issue on an Apple Watch Ultra 4, and I compared my Apple Health data before and after the transition to the new Watch/watchOS 27 on September 18. I excluded September 18 itself and compared equal 12-day windows: Sep 6–17 vs. Sep 19–30. The interesting part is that the number of HealthKit records barely changed: Samples/day: 43.4 before vs. 42.4 after Median interval between sample starts: 30:00 before vs. 27:22 after What changed dramatically was the duration represented by each sample: Median sample duration: 30:00 before vs. 3:01 after Mean temporal coverage/day: 20.69 hr before vs. 4.76 hr after Median temporal coverage/day: 22.23 hr before vs. 4.70 hr after So the issue does not appear to be primarily that HealthKit records are being created less often. Instead, each Environmental Sound Level record now represents a much shorter period of time. In my data, daily temporal coverage fell by roughly 77–79% while the number of records remained almost unchanged. I also checked a broader Aug 25–Sep 17 pre-transition period and saw the same pattern: ~30-minute median sample durations and ~18.9 hr/day average coverage before, versus 4.76 hr/day afterward. Apple Health shows my old Watch source ending at 10:26 AM on Sep 18 and the new Watch source beginning at 11:57 AM. The recording behavior changes precisely at that transition. One caveat: HealthKit does not expose the watchOS version on these samples, so my data alone cannot distinguish whether this is specifically a watchOS 27 regression, something related to the new Watch/pairing, or an interaction between them. But the behavior is very clear: the cadence of HealthKit records is nearly unchanged, while the duration represented by each record has collapsed from roughly 30 minutes to roughly 3 minutes.