Why? Probably because the reissue from start in a _1 copy was an indication that nothing was trusted with the client error code i.e. a stop not initiated by the servers. It's like running any other trickle-less result in that case, deliver 100% well or get nothing. Of course, logic would suggest that something could be salvaged, but as it is, the error/exit code is apparently ruling ATT, no process/rule to intercept the good parts.

(Vaguely remember at CPDN, trickles were immediately credited [daily update cycle or something], and error just led to a reissue, but don't remember whether that was partial [for the years not simulated, or whole, but think it was whole. Of course, if models that run for months on end crash and not giving credit would cause members to get seriously miffed, a reason I did stop as the ratio of success on my participating host was poor, and seriously no time to sit there doing backups and restores and what other micromanaging was needed... it required stopping 4 cores for longer, when 1 core only was used to run CPDN... for the truly hand-on)