motinis
commented
3 months ago As a note in terms of possible (?) fixes (copied out from Zulip thread):
This looks to me to be caused at this location in NS, where the prev. temp basal is used; my guess is that it's not properly cleared out when loop cancels things. Also note that the data there is sticky for the whole life of the temp-basal, the data that gets displayed for the temp-basal (see this function) uses a cache with a TTL, but the underlying data it will pull in after the TTL may not be refreshed. I'm really not an NS expert here; but possible fixes to me:
- Replace that prevBasalTreatment variable with an entry in the cache with a TTL
- Check the path upon which Loop updates the data; it should most likely result in a call to profile.clear() or something that will force an update. Maybe this is in profile.updateTreatments I think that prevBasalTreatment should be cleared
These aren't mutually exclusive.
Heiner1
This was seen using Loop.
Link to relevant Zulip message showing the issue: https://loop.zulipchat.com/#narrow/stream/144182-development/topic/Graph.2Fprediction.20discrepancy.3F/near/423206819
Seen with 0 basals, may occur with others as well (not known). When using Loop in auto-bolus mode, it will set a 0 temp basal. When Loop determines it is no longer needed, the temp basal is cancelled. Nightscout Basal pill (and basal display icicle) continues to display 0, while Loop pill states correctly that the temp basal was cancelled.
See this image (copied out from the above Zulip thread):
This has been seen on Nightscout 15.0.2.
Update March 14, 2024: Also have seen the other way around, where Loop enacted a 0 basal but NS is display a non-zero basal. Likely to be caused by the same underlying issue.