There is a lot of uncertainty with Resource Credits and how it will change as we move forward.
Please bear in mind, a lot of what I am going to write here is speculation. Some question are not answered and do necessitate looking into.
Here is what we do know:
- Networks cost money and there is a certain cost to running this network on a daily basis.
- Each day there is a certain number of transactions done.
- The cost of running the network does not change significantly from day-to day. Over time it will get more expensive as upgrades are required.
- There is a certain limit to the number of transactions that can be completed as the network is presently.
- Resource Credits are representative of the cost of running the network.
- Resource Credits are tied to the amount of Steem Power on a daily basis. Hence, the total fluctuates depending if more people are powering up or down.
I will use some of blocktivity's numbers even though they might not be accurate but they do show the point.
According to that site, we are doing about 860K transactions. This mirrors about what 
Blocktivty also says we are using .07% of the network. While I do not believe they have any idea of the network capacity, we do know the entire network can handle a lot more transactions. It did 2.5M at one point so we know a tripling is easily attained.
Thus, what is the number of daily transactions that could take place without major upgrades? I won't thousandfold it like Blocktivity claims but is 10 fold out of the realm? Could the network handle 8M transactions a day without taxing the infrastructure? I think that is fairly conservative.
The idea behind Resource Credits is to come up with an amount that most accurately determines what it costs to conduct each transaction on the network. We know this is a fluctuating number over time since not only will the number of transactions change but so will the kinds of transactions being conducted.
We obviously know the weight associated with each kind of transaction is also different with an upvote, as an example, being less than a comment.
This brings up a couple questions. If the total cost of the network is a known, to figure out the average transaction cost simply divide that by the number of transactions. Of course, to get the cost of each type of transactions requires further algorithmic breakdown.
The question is are the total costs on a daily basis divided by the number of transactions done in a day? Or is it divided by the total possible transactions? The method used is going to come up with two totally different numbers.
I am going to speculate that this move with the Hard Fork would not have been done using the total number of possible transactions. That does not properly reflect what is taking place. We know that most of the transaction potential, as expressed in Resource Credits, is unused. Opting for that method would drastically undervalue the cost of each transaction.
Therefore, I am going to conclude that the algorithm is programmed to reflect that activity that is actually taking place over a given period of time and uses that data to derive the pricing.
According to Steemd.com, the approximate costs for each transaction, excluding account sign up, is as follows:
This is where the major question comes in. What happens to these values change radically, in a short period of time.? For example, what if the transaction volume suddenly doubled in the next 24 hours maintaining the same category breakdown as it is now? The logical answer is the cost for each transaction would decline 50%.
I hypothesize this would be the case.
Taking it one step further, if a ten fold increase is possible on a daily basis without upgrading the network, wouldn't we see these same numbers drop 90%? Following the logic here, it should presuming most other factors remain fairly constant.
By the same token, a 50% reduction in activity, to 400K a day would, effectively, double the cost of each transaction conducted.
We need to mention there are other factors affecting this. If there were major moves in powering up or down, this affects the total Resource Credits. Also, the time frame that big shifts in activity, under these scenarios, would also have an impact. A 3 fold increase in a week is different than over 2 years.
Hence, could it be possible that we could see RC costs on the different transactions decline as more activity takes place? Even over time, the increase in network costs is not linear. In fact, upgrades tend to be marginal costs in many instances. Certainly, there comes a point where major upgrades are required at a certain volume level. However, before that point, increased could only require minor modifications i.e. changing out RAM as opposed to full server upgrades/additions.
The only question in all of this is the equilibrium the RC algorithm is seeking based upon the true amount of transaction conducted versus the potential?
If it is the former, which I believe it is, then we could see the RC needed for each transaction decline with more activity.
I will be interested if someone has the answers to some of these questions.
If you found this article informative, please give an upvote and resteem.