Any data set too big for one machine has to be split, and the split rule decides how painful growth is. The obvious rule, hash(key) mod N, spreads keys evenly but ties every key to the node count: change N from 8 to 9 and almost every key moves. For a cache that means a sudden cold start; for a database it means moving nearly all the data.
Consistent hashing places nodes and keys on the same ring and gives each key to the next node clockwise. Adding a node takes over only the keys in its new arc, about 1/N of them. Real systems give each machine many virtual positions on the ring so load stays even and a lost machine's keys spread over many survivors. Hashing does not fix a single hot key; that needs replication or splitting the key itself.
Choose a partitioning rule that keeps rebalancing small and load even.