Dexcom share bug?

[findthebug]

There is something strange going with the dexcom share API. This behavior is reproduced on pydexcom and my onwn implementaion. I get some duplicated entries back from the API. i also notices there is some new follow-onboarding in the Dexcom App on iOS.

They serve wrong/duplicated entries. best guess is the API is not working properly.

[gagebenne]

Please send code used to produce this output.

[findthebug]

its basically demo code from this repo.

user is "from europe registered by email", was working for ages now.

dexcom.py

i get same results in my own impelmetaion in swift. so i thinks is an API issue after all.

[gagebenne]

They are printing in reverse time order -- is that maybe causing the output to look "wrong"?

Some of the readings are closer together than I would expect -- they should be roughly 5 minutes apart. My readings are all pretty much 5 minutes apart.

[gagebenne]

How long have you noticed this duplication? Share (and thus pydexcom) just ferries and readings that your physical Dexcom broadcasts. There is a chance it is spitting out multiple readings, and that a sensor replacement may resolve things.

[findthebug]

2 days. i had issues with 2 sensors the same day. but now the new sensor is working as expected. beside that no issues. the strange part, in the official follow app i see the complete data as exprected.

[findthebug]

ok, maybe it is related to this:

https://www.dexcom.com/faqs/is-share-available-from-my-watch-when-in-direct-to-watch-mode

it looks like if you have appleWatch & iPhone, both sends data to the share service (in case you iphone is out of reach, there is a sync happening). this is exactly my setup and i got the info message today inside iphone about this new feature.

i keep my eyes peeled, and report back...

[findthebug]

Some new insights. I disalbed my watch for some time and then all goes back to normal. 5 min readings as usual. As soon as i turn on the apple watch, it messed up again.

1. 208 mg/dL (11.5 mmol/L) ↗ - 2026-02-15 11:50:27.095000+01:00 (double)
2. 208 mg/dL (11.5 mmol/L) ↗ - 2026-02-15 11:50:26.421000+01:00 (double)
3. 200 mg/dL (11.1 mmol/L) ↗ - 2026-02-15 11:45:26.455000+01:00
4. 186 mg/dL (10.3 mmol/L) ↗ - 2026-02-15 11:40:27.227000+01:00
5. 174 mg/dL (9.7 mmol/L) → - 2026-02-15 11:35:27.565000+01:00 (double)
6. 174 mg/dL (9.7 mmol/L) → - 2026-02-15 11:35:27.371000+01:00 (double)
7. 160 mg/dL (8.9 mmol/L) → - 2026-02-15 11:30:26.738000+01:00 -> Turned on Apple Watch & Breakfast :-)
8. 159 mg/dL (8.8 mmol/L) → - 2026-02-15 11:25:27.307000+01:00 -> Below here all good, watch was off for 2h
9. 157 mg/dL (8.7 mmol/L) → - 2026-02-15 11:20:27.010000+01:00
10. 153 mg/dL (8.5 mmol/L) → - 2026-02-15 11:15:26.784000+01:00
11. 145 mg/dL (8.0 mmol/L) ↘ - 2026-02-15 11:10:27.025000+01:00

I checked with proxyman and it looks like there are two service calls for share watch, and share mobile.

Thoughts:

It looks like the new share (watch&mobile) feature is buggy sending double values to the BE. They should just send one.

Or Dexcom does not care about doubles and they filter them in BE before pushing to the official share app. It makes kind of sense not checking sending doubles in the FE implementation because if you have multiple actors like watch & phone, its not easy to track this.

Since i just have one account, i can not test this with a second device & watch...

To fix this temporary in my iOS App, i will increase "max_count" for the same duration to get more values back. Not sure yet about filter for duplicates locally since i use this for charts...

If we can reproduce this on a second setup & dexcom allows duplicates, maybe an option filtering by seconds if values are the same, would be an option for pydexcom?

[gagebenne]

Ah, this is very interesting. Thanks a bunch for researching this.

I think I am going to leave the API as is. Leave it to library consumers to filter out duplicates, but keep the pydexcom library as simply a means of fetching the data available in the share. It just so happens that direct to watch users will get even more readings.