An orphan block reduces the DGM score without generating any payout, but a stale block doesn't, right? Since you are submitting blocks to multiple bitcoinds, the boundary between orphan and stale blocks is not that clear, when some bitcoinds can still accept a block while others would reject it. Since this distinction is affecting scores, it would be interesting to have more detail about how you decide whether a block that didn't pay was orphan or stale.
The distinction is pretty fine. It's considered a stale if the bitcoind that owns the private address that's paid out in the coinbase address never sees the block. It's an orphan if it sees the block and loses a race resulting in the bitcoin noting the payment as 'orphan'.Actually, I have wondered why pools don't modify their bitcoind to not reject a block, if the earlier block of same height still has zero confirmations. Why not work on the block the pool found, even if the other block of same height is already widely spread? If the pool happens to be lucky enough to mine a second block before the competing block gets a new confirmation, their earlier stale block combined with the new one would net the pool two blocks instead of one. Surely I'm missing some obvious reason not to do this.
Some pools are big enough to continue building on confirmation 1 blocks and win enough times to make it worthwhile. I think it's generally considered bad form by most pool operators to do this. There's a suspicion that some have done in the past though from discussions I've had with various operators. 
) and the pool finds a block that is later orphaned, the payments of the next valid block will be lower than they would be if the orphaned block never existed.