Everything has a dependency
The internet drops out. A controller restarts. A sensor battery goes flat. A service changes. None should be an unimaginable event in a system expected to be used for years.
I separate essential everyday functions from conveniences and ask what each one relies on. Local communication can reduce some dependencies, but it does not remove every point of failure.
Agree what remains available
For each important function, document the fallback. Can the light still be operated from a physical control? Can the heating be adjusted? Does a manual action depend on the same unavailable hub?
Check these behaviours in the selected system. A button on the wall is not proof that it will work independently.
Test recovery too
How does the house behave when power returns? What resumes automatically and what waits for a deliberate action? A successful restart should not create an unexpected scene in an occupied room.
Keep backups and a concise recovery guide where appropriate. Make ownership of accounts and ongoing support clear. Robust design includes the unglamorous moments.