I am tired of explaining this to junior developers who think they are architectural geniuses because they watched a ten minute YouTube video on microservices. Stop using completely random UUIDv4 as your clustered primary key on high-throughput tables. You are absolutely shredding your disk I/O because of page splits and index fragmentaton. It is not hard to grasp. If you absolutely must use a UUID for external references, use UUIDv7 or keep a sequential ID for the physical clustered index and put a unique index on the UUID. Refactoring a legacy system this week opened my eyes to how common this bad practice is, especially when they wondered why insertons were taking seconds. Please read a database book before you touch production code again.