Note that scattering can actually be beneficial if you have too much load for a single node when the data is hot. Of course you can solve this with weird sharding such as sharding on the least significant byte.
Seems like this point is not understood very well by many.
Ordered keys can create hot spots on insert and read, this can be good or bad depending on backend.
A hot-spot in one database index is collocated data in another. Sometimes you want uniform distribution of keys, sometimes you want clusters of keys because different databases are optimized based on assumptions about the underlying distribution of keys.
The problems are that they’re scattered during reads and writes, and will require updating old pages.