As the other fellow stated, single-page sequences are good for enforcing data dependencies between operations where there is no natural dependency. Ideally, you should see very few of them in code because there are very few cases in which there are trully no natural data dependencies.
A common mistake is to forget that errors are data too. As a matter of course I put error clusters on subVIs that can either generate errors, or might reasonably want to be able to respond to an error.
For example, take the case of the Wait(ms) function. It can't generate an error, but there are situations where it might need to be able to respond to an error. Say for instance the delay is very long (3-4 secs) if an error occurs in the code do you really want to delay for 3-4
seconds before you find out about it? Probibly not, that's why I rarely use the wait function by itself. Rather, I have created a routine that wraps around the function to give it error clusters and give it the ability to by pass the delay if there's an error input.
These error clusters also provide a natural data dependency reducing the number of artificial structures like one-frame sequences.
In terms of multi-frame sequences, you never really need them...
Mike..