Oh, one thing I didn't mentioned when I put in the BIP91 code on the pool:
I've changed it to always generate the 988,888 size blocks, instead of an 88,888 on the block change then 988,888 for all work until the next block change.
The reason is that 0.14.2 is 'supposed' to be faster at block transfers due to them finally implementing my idea from 5 years ago
(not transferring the full block, just the expected header/transaction needed, then whatever else is needed if requested)
Thus the time lost generating a full block 'might' be gained in the faster transfer of the block in or out on the net.
Hopefully they balance out, but I've not spent any effort comparing, since I'm not sure about how much is gained in the optimised block transfer.
If generation+transfer is slower, then that would mean a slightly higher chance of getting an orphan, if instead it balances out, then no expected difference.
Anyway, this covers the case before when we were getting smaller blocks if the previous network block was less than 30 seconds before our block.
Unfortunately, it would appear that the blocks are no longer as full as they were before, when the majority said SegWit was necessary ... sounds like a segwit scam to me
