I have a similar question.  The conclusion I'm coming to is that it's
implemented the way it is in order put the onus on the client - if the client
sends a large chunk of data and doesn't give it enough time to complete, oh
well... it's their foot; they shot themselves, etc.

If you have control over both sides of the wire for your implementation,
consider doing what I think I'm going to do: add an explicit
"requestedTimeout" parameter to the method (with a reasonable default), and
if you decide on the server that too much time is being requested, go ahead
and timeout/return on your own.

But, I'm all ears if anyone else has a better solution!