It turned out that this is needed for the mvme5500. The rather unintuitive
fact that the timeout needs to be longer for a faster CPU can be explained
by observing that the faster CPU can send more requests, thus gets more
responses, and response messages have higher priority (lower CAN object ids).