Practically, rollbacks due to transactions accessing an overlapping set of pages shouldn't be unexpected: the same client programmers who asked for BEGIN CONCUR

Practically, rollbacks due to transactions accessing an overlapping set of pages shouldn't be unexpected: the same client programmers who asked for BEGIN CONCURRENT are responsible for knowing what their queries do and dealing with their own conflicts by retrying the transaction or by avoiding or organizing concurrent access. Moreover, high performance applications that want to saturate disk writing rates can usually afford to organize their architecture around making massively parallel modifications interference-free (for example, support for the nonconsecutive key assignments the article suggests has vast ramifications).