RAID is only useful as long as you don't suffer more simultaneous disk failures than you've provisioned for (where "simultaneous" is dependent on your time-to-repair, since if it takes you a week to replace a dead drive, then two failures in the same week are indistinguishable from two failures in the same minute, in that you've lost the entire array).
That's why I suggested replacing at most 2 disks at the same time. But you could just one extra year and replace all of them one by one.
Or you can just use some dumb & cheap storage backend (in the cloud) as an extra backup layer, which you can easily replace at any moment if costs go wild or it goes bankrupt, without having to change the user experience.
Right. It's possible to stagger your replacements so as to avoid bad batches, or even to buy from different vendors for additional diversity (although for SSDs you have much more divergent performance characteristics), but when you're talking about rotating disks one at a time that's a lot of non-negligible overhead.
Depending on how much data you have to archive and frequency of access, it may very well be worth the cost to entirely offload opex and capex to cloud storage.
AWS Glacier Deep Archive in us-east-1 is $1/TB/month (GCP and Azure offer comparable pricing for archival storage). If you have 5x 4TB SSDs that, say, run you $200 each, and you run RS(5,3) then you're storing 12TB of data for $1000 of capex for 5-10 years. Meanwhile AWS would set you back $144/year, so the breakeven on capex alone would be if you would normally rotate your disks every 7 years.
Yes, but Glacier won't let you and your family browse easily and nicely photos and video safely stored, as Immich (or similars) does. It's a mix of Capex+opex+user experience+freedom of choice.
And yes, my wife regularly looks at photos from 3,4,5 years ago, or older.
Which is why I mentioned frequency of access. For backup purposes, archival storage might be good enough, but if you're looking for an alternative for frequently accessed content (along the vein of whatever Google Photos replacement) then most clouds will charge you at least an order of magnitude more (even Backblaze would use most of the entire proposed capex budget within a year).
But you might still have some value in using Glacier for archival purposes and having somewhat less resilient local copies (e.g. only being able to tolerate a single disk failure, or even just JBOD).
Either way, your current approach clearly works for your circumstances, even though it'd be too fiddly for me personally.
To be fair: my current approach is Immich saving photos on a single external USB SSD disk (1TB) plus daily backup on Backblaze (approx 5$/month). To maintain the same setup on Gphotos I would be paying radically more (obviously now I'm "paying" with the risk of a lower MTBF)
But I want to upgrade to a homemade multi-disks NAS just for the fun.
RAID is only useful as long as you don't suffer more simultaneous disk failures than you've provisioned for (where "simultaneous" is dependent on your time-to-repair, since if it takes you a week to replace a dead drive, then two failures in the same week are indistinguishable from two failures in the same minute, in that you've lost the entire array).