Voidscar Arena

Voidscar Arena +10, 2026-09-07 20:33 and 2026-09-09 20:58. Same dungeon, same key level, same +2 upgrade, same 30:00 par, one death each, 133 and 128 enemies killed. Hyporock's talent blob decodes to the same 78 entries in both, with an empty set difference in each direction. Item level 316.0 to 317.9.

The seasonal affix id differs (160 against 162) and the other four players are different people, which is the one loose end - but the other players cannot change his own damage per second, only the axis that ranks him against them.

  metric                     09-07 +10      09-09 +10
  damage done              253,748,181    320,636,758
  damage per second            206,853        253,384
  Holy Power spenders              468            516
  Blade of Justice / min          5.09           6.12
  Holy Power wasted              25.9%          26.4%

Per-spell, as damage per second, so the slightly different key lengths do not distort it. Only the movers are listed; they sum to the whole gain.

  spell                  09-07      09-09      change
  Divine Arbiter         7,248     28,705    +21,457
  Execution Sentence     9,610     15,517     +5,907
  Hammer of Wrath       15,846     21,172     +5,325
  Highlord's Judgment    9,085     14,247     +5,162
  Final Verdict         13,127     17,573     +4,446
  Light Within          13,030     17,059     +4,028
  Crusading Strikes     14,519     17,742     +3,223
  Blade of Justice       9,798     12,418     +2,620
  Divine Storm          48,590     46,401     -2,189
  Sun's Avatar          16,255     13,789     -2,466
  TOTAL                206,902    253,517    +46,615

Divine Arbiter alone is 46% of the gain. The rest is the shape you would expect from generating and spending more: the builders and the riders all rise together, and Divine Storm falls slightly because more of the spending went into Final Verdict.

The buff was never an uptime problem, and the intuitive read - keep it up - is exactly backwards. On 09-07 it was up 63.5% of the key and produced 118 damage events. This key it was up 24.8% of the key and produced 404.

The rule, established earlier the same afternoon from three nights of logs and not guessed here: Divine Arbiter is granted by a Holy Power spender, and the buff records WHICH one. It pays out on the other spender, once, and then it is gone. Grant it with Final Verdict and the payout is on Divine Storm, and the reverse. Holding it banks nothing and blocks the next proc.

A window counts as converted if the correct spender was pressed while it was up. Oldest first, Mythic+ only:

  key                       windows  converted  conv   median window
  Voidscar Arena +11             55         21   38%          19.2s
  Altar of Fangs +11             60         22   37%          20.0s
  Voidscar Arena +10             59         26   44%          18.2s
  Den of Nalorakk +15            73         20   27%          20.0s
  Voidscar Arena +13             69         25   36%          19.6s
  Ruby Life Pools +12            67         33   49%          17.7s
  The Blinding Vale +12          74         32   43%          18.5s
  Voidscar Arena +10 (this)     121        115   95%           2.4s

The median window length is the same fact from the other side. A window of 18 to 20 seconds is the buff running its full duration and expiring unspent; 2.4 seconds is it being cashed on the next press. Six of 121 windows went unconverted, against 33 of 59 on the identical key two days earlier.

Seven keys inside a 27-49% band, then one at 95%. The count of windows also doubled, 59 to 121, which is the same behaviour seen a third way: spending the proc is what lets the next one arrive.

For scale, the same measurement on a Ret Paladin - the second Retribution Paladin in this archive, 77 of 78 shared talents - runs 60%, 76% and 71% across the three keys he played alongside Hyporock. This key is above his best.

Four Retribution HUD layouts in the working directory were authored on 2026-09-09 between 17:56 and 20:30 - the last of them 28 minutes before this key started. They carry both Divine Arbiter buff ids and a last-cast gate that names which spender to press, and they prompt on Blade of Justice. This is the first Mythic+ key run with them live, so the two things the new layouts watch are the two things that moved.

Blade of Justice was the standing open item on this record: a tail of stretches over twenty seconds where the button simply stopped being pressed, against a same-spec control who had almost none. Same dungeon, same +10:

  Blade of Justice         09-07       09-09
  casts                      104         129
  per minute                5.09        6.12
  median gap               9.11s       8.59s
  gaps over 20s               11          10
  gaps over 30s                6           1
  time lost to them   16.1 s/min  11.8 s/min

The median barely moves, which was always true - when he plays the button he plays it at normal speed. The tail is what closed: six stretches over thirty seconds became one.

One honest caution on that half. Across his last five keys the time lost to long gaps falls steadily - 23.4, 19.3, 15.3, 18.5, 11.8 seconds per minute - but that series correlates with key level at r = 0.96, and this was the lowest key of the five. The like-for-like +10 pair is what carries it, not the trend. The Divine Arbiter half has no such problem: 95% against a 27-49% band is outside anything key level could explain.

Nothing here cost the key. The party stopped 42 of the 49 casts four interrupts could physically reach. Eleven of twenty-six Lava Bolts got through, which is the usual Voidscar tax on a pull offering 56 kickable casts to four kickers. One Poison Splash was left up for five seconds and did 392,322.

The death was Arcane Mage at 19:14 on Charonus: 779,856 of Void Cascade in four tenths of a second from 47% health. Nothing was reachable in that window and it cost five seconds on the timer.

The interrupt numbers above are larger than the ones this key was first graded with, and that is the interesting part. Hyporock landed ten interrupts in the log and the board credited eight. The two it dropped were **Mending Void**, the Voidminder's twenty-second heal on the Devouring Brutalizer - a CHANNEL, which logs a cast success the instant it begins and never logs a cast start. The board was built entirely out of cast starts, so the spell was invisible in both directions: eleven channels absent from the denominator, and every stop landed on them credited to nobody.

Fixed the same evening. Channels now enter the board from their cast success when the game data calls them channelled, with the channel's own duration as the window, and a caster killed part-way through is excused rather than counted as a leak. Re-graded, this key reads 42 of 56 rather than 32 of 45, and Mending Void turns out to have been handled well - **one of eleven leaked**, six stopped by an interrupt and four by a stun. Nobody's letter moved. What moved is that three players are now paid for work they were doing all along, and the highest-value kick in the dungeon is finally on the board.