Its not that i claim to throw manpower onto it regardless what the problem is. But without knowing what the bottleneck is i has to guess its something related to electric handwork. The software runs and is mining like we see, so it has to be physical work. And i dont think that its a so very special task that no worker form outside can do it.
No, it does
not "[have] to be physical work".
Here's two other possibilities.
1) For whatever reason, the software or firmware just isn't cutting it. E.g., one or more sets of hardware keeps dropping offline or getting 'hung'.
2) The chips / boards are failing nearly as fast as they are being deployed. Resulting in only about half or less of the first batch remaining operational.
Either of those hypothesis can explain why we're now seeing between 4 and 5th per hour rather than the 12 expected by now. In case one, they just can't keep them all cranking. In the second case, the magic smoke just keeps escaping.
I am not advancing either of these ideas as probable. However, with what we
know, either is just as plausible as your contention of what "has to be".
-- edit
Just went to take a look at current status. Now at "4,221.70 GH/s". Down from the about 5 TH/s we were at for a while. That is more than just mining variance. Clearly shows that not all the hashing power stays up once deployed.
