Increase active collator set size to 28

Hoping to start the discussion here.
It seems like a good idea to have this increase to 28 to have space for newer participants if so they wish or to invite those who are interested and have machines already running. We have seen a few showing interest around Discord so it makes some sense to do it. However, it seems that future increases will increase the gates to repeated identities and lower quality collators (maybe there already are a few there). So the increase to 28 seems like a logical final increase.

This is open for further discussion and maybe can be revisited in a few months or years, when we can have more interested parties so that a new expansion in the collator set can be reconsidered.

Hey Saxemberg, thanks for participating in governance and sharing your opinion. We need more of that to find the optimal decisions together.

Can you prove with data that repeated identities increase and that more "lower quality" collators enter the active set? What are metrics to determine that? How can we measure that?
It seems that you have observed interest for candidates to become collators, what would be an argument to cut them off?
Generally it is desired to push collator amount higher to increase decentralization, robustness and spreading of ZTG rewards, as long as it does not negatively impact system performance. Some of those goals are endangered by multiple identities of course. However, I find that multiple identities have an even higher impact in a smaller active collator set size, and the amount of ZTG required to enter does not change dramatically by adding new slots. We observed that required backing to enter the set lowers significantly after increments of the active collator set size, over the course of time the stake converges into the most promising candidates mostly though, effectively raising the lower ZTG boundary to enter the active set.
In the long run the stake should converge to the highest quality collators, and as of now this worked pretty well from my perspective, there are plenty of experts in the collator set.

@sea212
Can you prove with data that repeated identities increase and that more "lower quality" collators enter the active set? What are metrics to determine that? How can we measure that?
As for lower quality we have not had weird events, except for that one, where one collator degraded performance of the chain because they joined when they were still syncing. Other that that there have been no issues. However, we can imagine see issues like this happening more often if the threshold of entry is way too low.
As for repeated collators there are two things we can look at, the number of unnamed collators 3 in the active set and 5 in the inactive set. Which might or might not be repeated but, us collators, usually like to announce our identities or pseudo-identities. This is a completely subjective reasoning of course.
And a second important issue is the appearance of the first openly repeated collator at spot 35. We consider G-DOT a capable operator but reaching that spot 35 means that current collators will start fighting for such public second spot and future spots thus increasing the chance of altering the order or collators with less amounts staked currently in the active round.
For those reasons we have said that 28 is a good point to reconsider but we also believe that our original idea and proposal of 32 or even close to 40 is a good place to stop, something we don't have strong opinions against. We have to be mindful of the impact on the order of collators that are currently are active and that might have been active since the old testnets. But if Zeitgeist can manage a good balance between new identities, repeated identities and current collators then the expansion should be welcome as long as there is a plan to make all (or at least most) collators happy with the results. So far we believe, the results have been positive for everyone involved so all changes are open for debate and explanation.

@SAXEMBERG Thanks for elaborating the issues. In regards to the identities, AFAIK it is impossible to associate accounts to a real identity, i.e. someone could use a fake identity. I think this is a really hard problem and unsolvable. Even if we raise the awareness of the stakers in respect to the identities, the only effect I can see is that collators who want to run multiple nodes will setup fake identities. This is an immensely interesting topic (and it reminds me of my master's thesis, that tried to solve this fundamental problem "on-chain-identity<->real-identity" as groundwork for an advanced system), if anybody has any idea how we can prevent or reduce the probability of single entities obtaining multiple slots in the active collator slot, I am highly interested in your ideas.

I was told by a community member that it becomes difficult to stay profitable with increased reward dilution caused by increasingly opening collation slots. Can you confirm that? I think that alone would be a strong argument to stop increasing slots to avoid increasing pressure and competition between collators. Running into this situation would probably lead to congestion within a very small set of providers and probably even reduction of hardware performance below the minimum requirements to reduce cost.