This coin is too good to pass up solely because of these fork issues however annoying they might be. You guys will solve this eventually.
More info:
My "good" wallet that had staking disabled and I haven't mining with/to it also forked again at block 461516.
So now I know nodes can wander off to wrong forks even without any form of coin generation.
It basically stopped syncing and shows that it's 203 blocks behind but that's just keep increasing as blocks are being generated on the correct fork.
Checking the debug.log I have plenty of this:
received block 000000272a378005f92d
Misbehaving: 149.202.49.99:5757 (0 -> 100) DISCONNECTING
disconnecting node egc-seed5.granitecoin.com
ERROR: ProcessBlock() : block with too little proof-of-work
And a ton of these from even before the fork:
trying connection egc-seed1.granitecoin.com lastseen=0.0hrs
connected egc-seed1.granitecoin.com
send version message: version 70000, blocks=461605, us=0.0.0.0:0, them=51.254.216.25:5757, peer=51.254.216.25:5757
socket closed
disconnecting node egc-seed1.granitecoin.com
and:
trying connection egc-seed5.granitecoin.com lastseen=0.0hrs
connected egc-seed5.granitecoin.com
send version message: version 70000, blocks=461660, us=0.0.0.0:0, them=149.202.49.99:5757, peer=149.202.49.99:5757
socket recv error 10054
disconnecting node egc-seed5.granitecoin.com
These socket closings are suspicios to me. I don't have that for other coins and as I mentioned earlier these are happening even if I'm on the right fork.
Some more interesting clues, thank you for bringing them to attention. It would make sense that wallets that are not generating can wonder with peers, but it is a good 'process of elimination' fact to be known that you proved.
Initially I could not find the 10054 error code you mentioned. I then found that in log files of windows wallets. But it only appears the error codes must not be "standardized" across OS's as Linux wallets use code 104. The fact you have not seen this on other coins is, as you say, the worrisome fact. Possibly suggesting some underlying poor communication between nodes. Which may explain a lot here.
Again, thank you for the observations. I will continue to investigate.