A) Use existing open protocols: WTF do I need an app to get alerts. Doesn't email still work fine?
B) No cloud connectivity: No, camera, You don't need to store everything on the internet. My NAS can store it just fine. Or my backup usb drive, or whatever. If you want that as an option, sure, but quit trying to pretend that it's the only way this can be done. Same with you fridge. You don't need to store that grocery list online, it can just be emailed directly to me, TYVM.
C) Manual overrides: If my door won't unlock because servers are down, give me a physical key as well. I've read some companies faq's saying that if the internet is out, or their servers are down, you'll just have to call their customer service. No no no no no. Give me a key override. I'm not going to stand outside my own house while you try to get your servers going again. And what if your company goes under?
D) Open Source: If a bug comes out that renders my door-lock unusable, I want to know that I can patch it. We know the manufacturer sure isn't, using history as a guide. Why would you buy next years model if this years model didn't have security flaws?
Building a robust system isn't hard, it's just not profitable. All of these failure-points aren't there because we can't solve the problem, they exist because they are more profitable to leave unsolved.