Well the details of that problem at slush that was swept under the carpet, and slush hid the details ...
The problem was that he wasn't even checking statistics properly on the pool at least for the whole month of last December ... ...
For (just) that month of Dec last year at slush:
https://bitcointalk.org/index.php?topic=77000.msg13482822#msg13482822Pool found 230 blocks (2 stales) luck was 79.34870522%
So that means they would expect to find 1/(79.34870522/100) * 230 = 289.8598 blocks
So the pool was down (according to luck stats) by 58.9 blocks ... which was 1496.5BTC in December
Yes that was 1496BTC less than expected 100% that the pool got and paid to miners.
You can probably subtract about 1% for orphans.
The chances of being ~20% down on luck for 230 blocks is next to impossible
Any pool should know how many blocks a miner is (currently) expected to have found.
If you don't know that, you won't detect withholding.
If a miner is down 10 blocks it means something is already wrong.
Also chances of being down even ~10% are extremely unlikely on 230 blocks ...
So really we are talking at least 30 blocks missing that you can't account to bad luck.
750BTC
So what actually happened was Genesis Mining was running proxy software pointed at slush that they hadn't properly checked was working to find blocks on the live bitcoin network.
They should have solo mined to test their software, not made the pool miners pay for their testing.
What should have happened is Genesis Mining should have returned all the BTC the pool paid them since they did not (and would not) find any blocks.
But instead, slush decided that the miners should pay the price of Genesis Mining testing their software.
I wonder what deal slush made with Genesis Mining ...
Yeah slush may have the longest running pool, but it's clearly the longest running in the dark not knowing what he should have been doing about checking miners
Edit: I edited this years later due to putting the wrong name