Let me just get the next few rounds of trolling out of the way...
*puts on best kano impersonation*
The work restart will make the empty blocks that your pool abuses the network with less likely to happen*takes off kano hat* Not bad, eh?
But you still forget about the time the miner loses abandoning the old work, regardless of if it submits the already found results or not, it still takes an actual amount of time for the software and hardware to perform the full restart, which is lost time for the miner. This is one of the reasons different types of hardware have widely varying results on fast-restart type situations like p2pool since a huge percentage of their time could be spent doing nothing useful do to the restarts.
If the restarts were seemless then there should never be a dip in shares submitted when a new network block is found and all miners should just smooth into it like nothing happened. However, in reality, share submissions drop substantially at a full restart for some substantial amount of real time when it happens.
Heres some proof: Eligius block 301,134. Found in 12 seconds after the previous block using the "empty" template work. The pool wide hash rate average for the previous block: 5.8Ph/sec. The pool-wide hash-rate for the 12 seconds between blocks: 4.1Ph.
This happens for every fast block. A large percentage of miners haven't even caught up to the restart yet. Just look at
http://eligius.st/~wizkid057/newstats/blocks.php and look at the >9,000% luck blocks and the ones around them. Its a recurring theme: miners do not quickly update work and lose real mining time for every restart.
Funny how I have to keep repeating myself ...
...
Coz there is ONLY an insignificant amount of time lost during a full work restart on most hardware.
Most hardware supports a work restart at any time.
Some limited amount of hardware does NOT support a work restart at any time - and that SAME hardware has the same problem every single block change since that of course requires a ...... full work restart ........................
...
