There's a certain overlap here. It's not completely orthogonal. Having worked on safety critical systems before a lot of effort is put into detecting hardware errors in the software. E.g. random bit flips, ALU hardware issues, RAM writability issues, hash check of the loaded software being ok, plausibility check with (partually) redundant sensors.
You can detect a lot of hardware/QC issues on the software level. While it's still a hardware issue, better software can sometimes at least detect it
Excellent, to avoid killing a few people a year, you've killed thousands.
If you're not a diabetic or if you have no medical experience around this kind of device, kindly butt out and mind your own business. Low blood sugar in the middle of the night is an immediately deadly condition that needs treatment or the patient can end up with brain swelling. It's also not a condition that will wake the person experiencing it up. Having a CGM blare and alarm has saved countless people and given them a far better life from better sleep, less anxiety, and not randomly dying while resting.
Every CGM comes with directions telling you to calibrate the unit often and do blood stick tests to ensure the unit is working properly. Any diabetic should also be under the care of an endocrinologist as it's a complicated and deadly disease with lots of terrible ramifications.
It could be the software freedom conservancy assumed software bugs, with the same limited knowledge as the assumption being made here about chemistry quality control, so readers will have to decide which sounds more likely. The article do state later that "We also will probably never know whether this issue was in hardware or software... the public deserves to know the technical details ". We can make a favorable interpretation here that they acknowledge the possibility of it being software, hardware or QC. Making accident reports public information is a common step in other areas in order to allow people to learn from mistakes and produce better products.
I will add that blaming faults on human error has generally been shown to be a dangerous route when dealing with fatal accidents in all human endeavors. Correct training and behavior by patients can help to reduce fatal accidents, but one should always be careful to put blame here as a culture of blame generally produce more rather than less fatal accidents. Human-computer interaction is a complex subject and its very possible that the accident rate of those specific CGMs could have been reduced or prevented with better design, depending on what the issue actually was.