Another problem with the work-based target value.
If we have one work with a target value that has converged. And a second job gets added to the system which is equally lucrative, then it might happen that numerous miners switch to the new job. Let's say half of the miners switch, then the other remaining half in the first job will submit PoW submissions only at the half rate until the target value heals itself.
I more and more get the feeling that we need to have a global target value. Somehow!
Let us sort this out until not later than sunday ... So that we can start with the final remaining fixes.
If we have one work with a target value that has converged. And a second job gets added to the system which is equally lucrative, then it might happen that numerous miners switch to the new job. Let's say half of the miners switch, then the other remaining half in the first job will submit PoW submissions only at the half rate until the target value heals itself.
I more and more get the feeling that we need to have a global target value. Somehow!
Let us sort this out until not later than sunday ... So that we can start with the final remaining fixes.
If I recall correctly, the primary concern with a global target was that miners working on complex jobs would only get POW rewards a fraction of the time miners on easy jobs get rewards. This can be easily mitigated by the POW reward set by the author, (i.e. if you job is 10x harder than average, then your reward should be 10x larger than average).
Maybe for now we just proceed with this approach to keep things moving. And eventually (much later) enhance the UI to help the author set a better reward (i.e. recommend a POW reward based on historical averages, etc)
Ok, I'm going to say Jehova here:
Why is it so bad to have the network mine on one job at a time? A Job gets processed by the whole network, then the next, then the next? If we either set a fixed PoW reward, or we could implement a function for miners to reject a job, and if enough miners reject a job, it gets terminated, I don't see that much of a problem. I mean, if we have a fifth of the network each working on five different jobs, or if those five jobs get worked on one after another is in a general sense not important, is it? If we want job authors with big funds to "cut line", maybe they can buy a premium spot in the line by having higher PoW rewards?
. 
