Do a staggered backup
Assuming you are hassos on a pi with sd card (inferred from posts), perform scheduled configuration and image backups to both a usb flash drive as well a local network location
Backup the local network location in an encrypted form to a cloud storage provider daily (when it detects changes obviously)
Perform a full image backup prior to any system upgrade in case it goes wrong
Document the backup strategy AND recovery process in great detail so if something does go wrong, you do not need to start from scratch figuring out how to get your system working again.
Monitor the backup locations on a regular basis, make sure you can get the files if you need it, test cloud storage downloads for transfer speed and file integrity.
Using another sd card, TEST YOUR RECOVERY STRATEGY SEVERAL TIMES A YEAR
I would STRONGLY consider moving your bootable storage to a SSD, and only use the SD card (a high endurance card) for local backups, and using a flash drive for recovery testing.
The reliability of a system like this falls squarely in both the “you get what you pay for” and “you reap what you sow” categories, inexpensive off the shelf sd cards and hobby grade boards like the pi are not built with reliability and redundancy in mind, no matter how reliable the software running on them happens to be.
Use “enterprise” or high reliability storage devices, get a BIG battery backup AND a high end surge protector, use RAID on networked storage, and backup everything everywhere.
It can be run an many devices, I am running it in Docker on a basement server that performs many functions, it is fast, so that also makes backup and recovery fast, the primary backup location is on the same server different SSD. That location uses Duplicati to backup encrypted to Backblaze B2.
I can stop HA, backup all my config and a 3GB database, upgrade to a new version, and have it running again in about 4 minutes.
Recovery from a backup file is even faster, stop HA, rename the corrupt config folder, unzip the backup to the correct location, and startup of HA is well under a minute.
That is the kind of recovery system it sounds like you might want, but it requires maintenance outside of HA and hassos, since you are now in charge of maintaining the host operating system on which Docker is running. It is also FAR more expensive, the server, with 20TB storage, 32GB of mem, and 4 solid state drives (3 are in RAID5) cost about $2500.
There are of course far less expensive intermediate solutions, many use a NUC or other small x64 mini-pc, they run from $300 to $1000 configured with varying levels of performance and reliability, though they do not have the kind of hardware redundancy options that an actual server has, you cannot have a RAID6 storage array for example, or a redundant power supply.
The question to ask is “how much does it feel like I lost” if your system goes down and is unrecoverable, the hardware cost of course, but also the work you put into it. If you value your own time at say, $20 an hour, and you spent 100 hours setting it all up, that 2 grand right there! Spending a fraction of that to make the chance of another loss far less likely and to also make the recovery from a loss far easier and more rapid is a no-brainer… if you have the money to spend.