I've just made a slight change to one of the factors we use for calculating balance, so everyone's balance times should drop down by about 0.1 seconds.
Please keep in mind that, with this new system without the quarter second rule, some balance times will increase slightly, while others will decrease.
However, also keep in mind that the previous balance times that were reported to you with the config option were not accurate. They were as accurate as we could provide at the time, but they did not represent what was actually happening.
Examples:
Under the old system, an attack that was calculated to use 3.6 seconds of balance would be scheduled to use 4.00 seconds of balance. In practice, you would recover balance, on average, 3.88 seconds later. At best, you would gain balance after 3.75 seconds, and at worse, you would take the full 4.00 seconds.
Under the new system, an attack that is calculated to use 3.6 seconds of balance will be scheduled and reported to use 3.80 seconds of balance. In practice, you will receover balance 3.80 seconds later (with maybe a little bit of difference due to CPU time).
Under the old system, an attack that was calculated to use 5 seconds of balance would be scheduled and reported to use 5.25 seconds of balance. In practice, you would recover balance, on average, 5.13 seconds later. At best, you would gain balance after 5.00 seconds, and at worse, you would take the full 5.25 seconds.
Under the new system, an attack that is calculated to use 5 seconds of balance will be scheduled and reported to use 5.2 seconds of balance. In practice, you will recover balance 5.2 seconds later.
As you can see from my examples, there are situations in which the new system will decrease balance times, and others where the new system will increase balance times. This is due to the fact that the previous system had both a very low level of granularity mixed with a significant random factor.
The new system is much more precise, and on average across the whole range of possible values for balance time, the end result is that the actual average time it takes to regain balance is consistently similar between both the new system and the old system.
I apologize to those of you who had your balance times slightly increased by this new system. It is, unfortunately, unavoidable in order to move to this much more accurate system. Please keep in mind that we are talking about small fractions of a second here, so the actual effect when taking latency into account is negligable.
Penned by my hand on the 11th of Letum, in the year 635 AD.