It’s probably a Hall effect switch that turns on when the magnet is removed. The power management chip is then built so that it keeps itself on once power is available, meaning that replacing the magnetic won’t turn it off. That makes restarting much more difficult - you have to open the case and physically remove and reinsert (or replace) the battery to get everything to reset.
I didn’t check the pictures in detail but most of the time modern IoT things have either Flash or data (CMOS) in the clock which allows them to store data permanently.
My suspicion is that the unit stays switched on so that when we do a return Dexcom can simply scan the device without having to put it into any special equipment. Their “limited replacement” policy seems to depend on being able to track the returned devices.
Flash is semi-permanent (it should last a decade without power). So called “CMOS” (almost everything electronic is CMOS these days, and much of it is not memory) actually depends on the battery in the clock chip to keep it alive. My guess is that the G7 doesn’t include a separate battery for the clock, so any CMOS memory will go away when the battery is dead or if the battery is removed. If the G7 shuts everything else down at the end of 10.5 days, the battery will probably still have enough energy to keep any CMOS memory alive for at least months, possibly a lot more, so there is plenty of time to return a defective unit.
Try twisting the two plots around their common point at about 1:10PM. When I do it I see what I term a “drop out” on the old sensor at about 12:30PM but then a very good correlation.
I rather suspect the new sensor (the one still giving readings at ‘Now’, I assume) is inaccurate in absolute terms.
I think some people reasonably say that calibration is required but I’ve had problems with that on the G6 so I don’t do it; I just use the overlap to allow the G7 to come to adjust to my body as best it can.
I have a big problem with accuracy of G7 the first 12-18 hours after insertion. I started putting them in during the grace period hoping that would help. You can see the one in this photo reading below 55 when the active sensor which hadn’t expired yet had me in a normal range. This inaccuracy continued after I started the new sensor. Required many mini calibrations to get it to read correctly. Last insertion the issue was it was reading 70 points higher. I came to home from the gym and it was reading 145 and I was 64. I am aware of lag time. This is not the problem. I’m really fed up. Anyone else have this problem. Pre soaking the g6 improved first day accuracy. Not g7
I don’t trust my cgm (Dexcom) the first day. It is always off. I want to calibrate to make it more accurate, but have been told and I read not to do that
You could certainly try calibrating. Some people seem to find it useful. My experience is that sensors grow more accurate over those first 24 hours on their own.
Thursday night I put in a new sensor. Yesterday morning (Friday), when I connected the new sensor to the app, it was reading low. Feeling impatient, I put in a calibration while my BG was steady. By the end of the day, the G7 was reading higher than my fingerpick BG so I calibrated some more. At this point it’s doing fine, but I personally wish I hadn’t calibrated. I think it might have been more accurate. Also, I don’t feel comfortable with a CGM that reads high as it can trigger unnecessary corrections and increased basal rates.
Overall, it wasn’t a terrible experiment but it cemented my preference for no calibrations in the first 24 hours even when the readings are off.
My experience has always been that if I insert a new sensor when I am at or near my ideal BG, then corrections are never needed. I have not calibrated a sensor in the past 4-5 years. So my normal ideal BG is 100 mg/dl and I insert my sensors when my BG is between 100-110 mg/dl. If I insert when I hit 150 or above, it will take several hours to stabilize.
That’s what I feel with both the G6 and the G7; their readings are complete shyte for, it varies, maybe 6-24 hours but after that they are sort-of fine.
When I calibrated the G6 was shyte til it died.
The G7 may have better recovery from a calibration; I’ve done it I think two times when I couldn’t do the overlap and it was clearly outside the ballpark.
The overlap is best (and I’ve now discovered how to do it with xDrip+ and get both sets of readings, not that it is easy 
I now calibrate when my glucose has been level but the Dexcom G7 is off. I will calibrate immediately if I get crazy low readings. It seems to get rid of the stupid first 12 hour readings I would get presoaking
