-
As to how they break: Their memory can only take a number of writes before it fails... enough bits failing and the SSD can no longer find places to put files. So they degrade whilst bits are failing
This seems to suggest the total capacity of the drive would "shrink" the more it's being used.
I always thought they retain full capacity ..but then just go die at some point and that's that.
salad-
They retain their full readable capacity... but in most places they are used the data changes.
i.e. within a cache layer the algorithm applied might be LRU (Least Recently Used), which means to put new items into the cache the oldest items are evicted. Eviction here just means over-written... but if you can't write, then for the purpose of cache your storage is reduced as you can't add new items to the cache.
Example on LFGSS... the attachments. The attachments live in an LRU cache, and new attachments in hot (most visited) conversations will be requested the most. So those items are added to cache, and because cache is always full (it's expensive SSDs and less capacity than long term storage) the least recently served (oldest) item in cache is deleted from cache to make space for the new item. If I can never delete the oldest, then eventually I can never add newest, making the cache server useless to me (because requests would be cache misses and reach the far slower spinning disks somewhere... I care about speed, as slower reads take longer, resulting in contention, which floods networks - and networks are the problems cloud providers really have. Fast caches, SSDs, save networks.
Velocio
It's not really possible to recycle them. They all get shipped eventually to parts of India and China where low paid labour just pulls chips off of the circuit boards for classification, and these are later sold at a fraction of the cost.
By and large though... no significant recycling happens to any of the tech that the human race produces, it all goes to landfill. The cost of recycling being greater than the cost of the item to produce again. All of the efficiency that exists, exists in the production of these things at a low cost with acceptable pollution and externalised costs... there's virtually no efficiency in resource re-use and preservation, recycling, etc. If you want a capitalist society with a liquid economy and to be able to afford most of what you consume and purchase, this is what that takes today.
As to how they break: Their memory can only take a number of writes before it fails... enough bits failing and the SSD can no longer find places to put files. So they degrade whilst bits are failing, and once enough bits have failed they degrade faster as fewer bits now receive all of the writes. As soon as the OS can no longer write to an SSD the drive is pulled and discarded and a new one added. If multiple SSDs in a server died (replication exists within a single server too) then the server itself is removed from the fleet (networking stops sending traffic to it) and all of the SSDs are replaced (if the server is a high cost one) or the server is just permanently routed around until the vast majority of its' rack has died - and then the whole rack is replaced at once (a cheaper operation if your fleet is cheap commodity hardware and you have a lot of it).