Does clustered index fragmentation happen in relational DBMS?

Is it possible that after a certain pattern of random inserts and deletes **leaf data nodes** become fragmented with **clustered** index? I.e, that the physical order does not reflect the logical order (by say, INT primary key) imposed by the clustered index? This way, range queries would require random I/O even after finding the beginning of the interval. Most university courses (ex. CMU Introduction to Database Systems by Andy Pavlo) say that data is physically ordered according to the key. While definitely approximate to the reality, that looks as unrealistic to me counting the cost of frequent file defragmentation that would be required.