A dead VPS is less painful if the setup assumes it will happen. Some design choices that help:
The VPS runs MT5 plus a tiny script that writes a heartbeat file every minute.
A second machine, home PC or a cheap cloud function, checks the heartbeat and sends a phone alert if it stops.
Every order carries its stop and target on the server, so nothing depends on the terminal staying alive.
EA settings and templates live in a backup folder that can be copied to a fresh VPS quickly.
A written note lists exactly what runs where.
A test worth doing quarterly: shut the VPS down on purpose on a quiet afternoon with a micro position open, and time how long it takes to notice and recover. If it's longer than you'd like, the alert or the backup isn't good enough.
One subtle risk: two terminals running the same EA after a recovery, both trading. A simple guard is a magic number check plus a lock file, so only one instance acts.
My VPS died at the worst possible time once, during a quiet afternoon when I had wandered away from the desk, and I did not find out for nearly three hours. The positions had stops on the server so the damage was limited, but the EA missed two exits it would normally manage by trailing, and one trade gave back most of its profit.
After that I set up something very close to the heartbeat idea. A small script writes a timestamp to a file, and a free scheduled job on another machine checks it and sends a message to my phone if it has not updated for five minutes. It cost me an evening to set up and has caught two outages since.
The duplicate terminal problem is real, too. When I rebuilt on a new server, I briefly had both machines running, and they both opened the same EURUSD trade. Now the first step of any recovery is confirming the old instance is fully shut down.