We manage a system where there are thousands of reads for every write operation so we use clustered indexes on the keys used to do the majority of joins (not always the primary key). Performance is woeful if we rely on normal indexes for these key join relationships. We also have the added issue that the majority of key values in the system carry additional meaning i.e. the range of numbers (INT) also implies business grouping ownership which means clustering also has a side effect of storing all the groups in an order - rather than being a burden, this happenstance appears to offer many benefits