My bonuspool account went apeshit exactly like that for a little while a few hours ago. I switched a couple parts around in my rigs so I wasn't sure if it was a hardware issue, but it seems to have gone away.
I don't think it's bonuspool. The common element was that we both had a pool go down. I use Eclipse and mtred, and when Eclipse went down this morning, I manually added mtred because it wasn't in the config at the time. Later on, I realized Eclipse had come up and was up for awhile, but the miner was doing -exactly- what Morblias was saying: except that I switched it myself, and it was failing to switch back. Actually, I didn't even switch it - I just added mtred manually while it was running, and it went over to mtred on it's own. Of course at first I expected the errors it was giving... because the pool was down. I even checked using ping.eu's port checker function and 8337 on eclipse was closed for all Eclipse servers. Later on when I knew the pool was up, the miner was doing exactly the same thing. Sitting on mtred, and saying Eclipse was down. I even checked again with ping.eu to make sure, and indeed Eclipse was up. So I just closed the miner, repoened it, and voila, connection to Eclipse.
I think this is pretty solid evidence
I can confirm similar behavior as well in yet another scenario. I added a 5th pool (I already had pools in positions 0 through 4) through the curses interface. I then proceeded to switch to it. Shortly thereafter, cgminer switched back to pool 0 (presumably due to a problem with pool 4), and then I started getting alive and down messages every few minutes but not submitting shares to pool 4 (I also started submitting shares to pools 1 and 2 álong with 0 in spite of --failover-only and 18 hours since launch of cgminer). I deleted pool 4, added it again, switched to it again, and voila, cgminer is mining on it again. Next time I may try disabling and re-enabling it instead of deleting it, however I can also report that it said disabled on the pool after I hit P before I deleted it. I thought they were only supposed to go to disabled (vs dead) after a long string of rejects, and I didn't see any rejects. I don't know if this additional information is helpful or not, but something is up...
ETA: I am using 2.7.1, I think the two quotes above are too, I just realized none of them indicate it in this post.
