Rather than making one node that never goes offline, which is very hard to do, it's better to divide the staking process among at least 3 nodes in separate locations. One can be in a datacenter, one at home, one at eg. parent's home.
If each node is online for 99.9% of the time (8h 45m of down time per year), the probability that at least 2 nodes are online at once is 99.9997% (2m per year) - assuming the causes are independent.

In addition to safety from being offline, safety from slashing due to hardware errors is gained: only two nodes at once can generate a valid block. So even if one node has some weird hardware error that causes it to forget it voted and it generates a slashing offense, almost certainly the error isn't going to happen at the same time elsewhere.

The code is very simple and adapting it to staking should be trivial.