This 2009 reminiscence discusses the term’s use in the UO context in 1996, so 2003-2005 mentions don’t affect the author’s conjecture at all. (I’d guess there

This 2009 reminiscence discusses the term’s use in the UO context in 1996, so 2003-2005 mentions don’t affect the author’s conjecture at all. (I’d guess there are uses of ‘sharding’/‘shards’ in this sense from much earlier, perhaps decades earlier, database/caching work, but I could be wrong.)