Smart-home automation is best suited to repetitive, reversible, low-consequence tasks; functions involving safety, access, water, heat, vehicles or essential equipment deserve stronger manual fallbacks and more conservative design.
Set up the network first, inventory the devices second and rebuild automations last. That sequence reduces rework and makes it easier to separate Wi-Fi problems, account problems and device-specific failures.
A Home Assistant smart home can keep local lights, switches, sensors and automations working during an internet outage when the devices and dependencies remain local and powered.
A small x86 mini PC can be an effective always-on smart-home controller, either as a dedicated Home Assistant OS appliance or as a Linux host running Home Assistant in a supported VM or container.
A cloud-independent smart home starts with locally controllable devices and a local controller such as Home Assistant, because software cannot remove a vendor-cloud dependency built into the device.
Dead Internet Theory: Repeated smart-home events can reveal patterns of occupancy and daily activity even when no single sensor reading seems especially revealing on its own.