LOL
It seems to be going the wrong way back up to over 200 pages again I just checked and it was 205 pages waiting for validation and 6 pages waiting for verification so 3165 WU backlog. What is normal it appears that just looking at the returns with todays date on them I have turned in 23 pages that have been validated and 14 pages that are waiting to be validated so about 38% of daily work in waiting. This is just a small sample of what my daily WU count is so it may not be that accurate but it should be fairly close.

My PV hiked quite a bit since last few days after the major drop when repair deadlines were dropped to 20% of original deadline. They're now back up to 35%, but since MCM has now an original deadline of 7 days the expectation is that batch completion will drop [for MCM] from 16 days to 12 days]. The change effects will take a few days to trickle through.

The smaller the cache the bigger the PV, since wingman matching does not exist. Those with small cache are usually first to return the result and then, with a mean return time of about 3 days, those running with a bigger will report much later.

My PV ratio is currently about 85% of a daily completions. That's with a 1 day buffer. If grandpa thusly had a 1 day buffer and does 4000 a day, the PV total at any time would be around 3400 [very ballpark].

It gets worse when the weekend starts... all those office machines switched off [doing 8 to 5 on M-F days] with incomplete/unreported work on them. Work fetched in the weekend usually has wingman that also compute in the weekend, so by Sunday late/Monday/Tuesday [usually peak global stats day] you'll usually see a drop again, more so when the office machines are powered up again.

At any rate, we have many threads on this topic. The shorter the deadline, the smaller the PV. Projects have to strike a balance between 24/7 crunchers, storage issues and part-time crunchers, those that do 1-2 results in a day/week. You don't want to lock those out. Summation, we have to accept a certain PV ratio. If ever the scheduler would have the capacity to match wingman based on return history, maybe the number of PV would drop. Is it worth the effort if the overall PV sum does not change? No, don't think so... just live with the idea that credit will always arrive at some point for valid results... usually for >90% of the results within 4 days [old stats info data point from techs].

If only a single batch of wu's is important, needing to wait 12 (or 16) days on the results of the batch would be very long if enough results is finished per day to equal a full batch but the results is spread-across different batches no batch finished in a single day.

If on the other hand finishing all batches is important, things would be different.

Let's make an easy example, by assuming enough crunching is done to equal 1 batch/day and the total number of batches to crunch is 1000. Let's also assume with a 7-day deadline it takes 12 days to complete an individual batch.

Let's also assume, if the deadline is decreased to 6 days, it takes 8 days to complete an individual batch. Also let's assume 99 % of computers uses less than 6 days to return a result, while the remaining 1 % uses between 6 and 7 days.

To finish project with 7-day deadline:
1000 days crunching + 12 days delay at the end: 1012 days total.

To finish project with 6-day deadline:
1010 days crunching + 8 days delay at the end: 1018 days total.

Meaning, even a 1 % decrease in production due to shorter deadline will more than cancel for faster turnaround-times.

While only 6 days extra isn't much, this was with an assumed 1 % decrease. If instead the decrease is 10 %, the total time jumps to 1108 days and over 3 months additional time will be significant.


BTW, if individual batches does have some useful scientific results on their own and it's not only the total what's important, a 12-day delay seems long but this is only until the 1st. batch is finished. Afterwards, a continuous trickle of batches will be finished and can be handled more or less daily.

Not sure if example a 100-generation FAH-Project can give some useful information from a single run/clone before all 100 generations has been finished or not...