New message for subset of days #2073

Activity

1. bartmuskala

   added

   bugSomething isn't workingSomething isn't working

   on Sep 10, 2026
2. ryanbr

   added 4 commits that reference this issue on Sep 11, 2026

   Say WHY a night missed the re-score cache, not just how many did

   ...

   6fa8fb1

   Say WHY a night missed the re-score cache, not just how many did

   ...

   2b1af0c

   Say WHY a night missed the re-score cache, not just how many did

   ...

   9e11004

   Say why a night missed the re-score cache, not just how many did (#2080)

   ...

   3812131
3. ryanbr commented on Sep 11, 2026

   

   ryanbr

   on Sep 11, 2026

   Owner

   More actions

   Thanks @bartmuskala. The log answered it, and the short version is that the message is working as designed, but you have lost something real and it is worth saying exactly what.

   What happened. Your Charge stopped scoring the moment the app restarted at 18:36 on the 10th. The same 21 nights, the same data, only the running build changed:

   18:33:39  charge=19 effort=19 sleep=19 hrv=19 rhr=19
     ===== session rolled at 16:36:22Z (18:36 local) =====
   21:48:43  charge=0  effort=19 sleep=19 hrv=0  rhr=19
   

   Charge went with HRV, and HRV went because of a deliberate change. WHOOP 5 interval words turned out to be 1/1024-second ticks rather than milliseconds, so every historical interval had been inflated by 2.4 percent. Rows recorded before that was known mix correctly-converted values with wrongly-scaled ones and nothing stored separates them, so NOOP now declines to score that mixture rather than score it wrongly. The note you are seeing says so. Sleep, Effort and resting HR are unaffected because they read heart rate, not intervals.

   What it costs you, plainly. Nights recorded from now on score normally. Your existing nights will not come back, because the strap has already trimmed those records and there is nothing left to re-offload. That is a genuine loss of three weeks of Charge and I would rather state it than let you discover it.

   You also found something separate, and larger. Your log shows the nightly re-score taking 193 seconds and reading 1.75 million heart-rate rows, every 15 minutes, on battery. An earlier log of yours did the same 21 nights in 16.5 seconds. The day cache that exists to prevent exactly that reused nothing on 13 of 17 passes, while the steps cache in the same passes reused 58 to 60 of 60, so it is not general slowness.

   I could not tell from the log WHY it reused nothing: the line reported the count and never the cause. Three changes now fix that, and they are diagnostics only, nothing about behaviour:

   • PR Say why the Rest card is pending sync, instead of changing in silence #2077 says why the "Pending sync" state turned on or off, which is the other thing you reported
   • PR Make two export headers say what they actually mean #2078 stops two export headers stating less than they mean
   • PR Say why a night missed the re-score cache, not just how many did #2080 says which part of the cache key moved, so "reused nothing" carries a reason

   What would help. Once the next testing build is up, one more strap log taken while the app is behaving like this would very likely settle the performance problem outright. No rush, and nothing to do differently, just the export.

   Sorry about the Charge. The units fix was right and the evidence behind it is solid, but the blast radius on an existing install is worse than it should be, and that part is on us rather than on your strap.
4. bartmuskala commented on Sep 11, 2026

   

   bartmuskala

   on Sep 11, 2026

   Author

   More actions

   Thanks. Unfortunate to hear.
   Wouldn't it be possible to have a fallback in place? Charge scores were potentially not correct but couldn't be completely off? The baseline should be different but high would remain high, a low score low?

   I prefer to have A score with a message explaining the background/nature of the different calculation, than 'No data'.
5. bartmuskala commented on Sep 11, 2026

   

   bartmuskala

   on Sep 11, 2026

   Author

   More actions

   My log

   noop-strap-log-260911-0802.txt