Voidscar Arena

Nobody in the party played badly enough to explain twelve deaths.

The tank took zero deaths and graded A. The healer graded B with response 96 — 'response' meaning output actually rose during the party's worst damage windows, rather than absolute healing per second. Both are near the top of what the scale expects.

What happened is a route decision and an interrupt assignment. Two pulls carried 42% of the entire run's damage, and the mob generating most of the unanswerable damage in both of them is one that better groups do not fight. That is a fixable problem and it is not a personal one.

Time ledger for the run: 1,101.5 seconds total, of which 1,018.6 was combat and only 57.5 was routing. The group spent 92% of the key fighting. Whatever went wrong, it was not walking too slowly or taking a long path.

  #   start   length  enemies   damage taken   deaths   notes
  1   0:09     85s      12       21,380,182      0      17 of 35 stoppable casts stopped
  2   1:40     76s       7       14,938,855      0      Raj'kess the Spellstorm
  3   3:02    279s      25       60,764,145      3
  4   7:50    186s       -       52,171,115      0      Taz'Rah, killed
  5  11:08    143s      21       43,364,409      3      two Agitated Voidscythes
  6  13:36    180s       9       33,791,868      2      one Agitated Voidscythe
  7  16:54     71s       1       19,111,878      4      Atroxus, wiped

Pulls 3 and 5 together account for 104,128,554 of the run's 245,534,192 damage taken — 42% — and six of the twelve deaths.

Pull 3 is the slowest pull in the key by a distance: 25 enemies in 279 seconds is 11.2 seconds an enemy, against 6.8 for pull 5 and 7.1 for pull 1. It is a four-and-a-half-minute engagement that took a quarter of the run's total damage and killed three people. Whatever it chains together, it is too much at once.

Pull 1 is a separate problem. It leaked 18 of the run's 21 unstopped casts — 17 of 35 stoppable casts stopped, against 12 of 15 in pull 5 and 5 of 5 in pull 6. It is also where the party bottomed out: Doryhunky read 11.9% health at 0:50, and the healer's entire emergency kit went out between 0:20 and 1:01. That opener was not a panic dump into nothing; the damage was real. But it means the party played minutes 1 through 6 with no big externals available.

Four separate Agitated Voidscythes were engaged, in three fights. None were summoned; they are ordinary pull mobs.

  #1  11:14 - 12:12  killed
  #2  12:32 - 13:16  killed   <- two alive
  #3  12:32 - 13:30  killed   <- at once
  #4  14:39 - 15:29  killed

One Voidscythe casts Corrosive Essence roughly every 21 seconds and hits three people. Two overlapping produced five-target waves at 12:40 and at 13:01. In the 30 seconds around 13:01 the party took 32,003,554 damage — the densest stretch of the key by a wide margin. Hyporock died at 13:11, Shadow Priest at 13:26, Arms Warrior at 13:28.

On the comparison side, the reliable finding is a negative one: searching the full page text of all four top logs returns zero occurrences of 'Voidscythe' and zero of 'Corrosive', across rosters of 65 to 84 enemy types.

A partial positive roster from the +20 lists Charonus and Aegyra the Unyielding, neither of which appears anywhere in this run. That list was only partly extractable and should be treated as a lead rather than a fact — and this run was abandoned before finishing, so Charonus may simply be a boss the group had not reached yet.

Corrosive Essence has a cast bar of roughly 2.5 seconds — the log carries a SPELL_CAST_START and a SPELL_CAST_SUCCESS about that far apart for all eleven casts.

From the 12.1 client spell data (wago.tools, table SpellInterrupts):

  Lava Bolt         1228176   flags 0x2d   interrupted 9 times this key
  Mad Shriek        1233398   flags 0x2d   interrupted 10 times
  Shadowbolt Volley 1299938   flags 0x29   interrupted 6 times
  Corrosive Essence 1289265   flags 0x29   interrupted 0 times

Corrosive Essence carries the same interrupt flags as Shadowbolt Volley, which this party kicked six times in this very run. That is strong evidence it is kickable, not proof — but the party's 37 interrupts went to Lava Bolt and Mad Shriek while the spell doing 20.9M went out unstopped eleven times.

It is also dispellable, and the existing cleanse prompt already covers it. But a Cleanse on roughly eight seconds cannot answer a debuff landing on five people inside four seconds. Stopping the cast is the only response that scales; cleansing is the cleanup.

A (89.8), and the only member of the party who did not die once in a key where the other four died twelve times between them.

  survival    100   0 deaths (deducted, not averaged — no charge here)
  mitigation   92   99% uptime of the 100% the kit can hold; 100% of worst windows covered
  interrupts   91   20 stops against 15 expected of him
  activity     96   98% engaged while alive in combat
  damage       85   116,092 dps, 66% of the average damage dealer
  utility      83   1 used of roughly 3 the kit offered; Imprison never pressed
  mechanics   100   0% of intake was damage comparable players did not take

He absorbed 99,819,296 of the party's 245,534,192 damage taken — 41% — which is the job, and he did it without dying. He also carried the interrupts personally: 14 kicks plus 6 crowd-control stops of the group's 37 total. Three of his five biggest damage targets were the packs that hurt most, including 11,230,587 into Agitated Voidscythes.

Three entries the report excuses rather than charges him for, because the enemy chose the targets: Rip and Slice 65 hits for 6,842,453, Nether Dash 86 hits for 4,469,796, Corrosive Essence 26 hits for 2,881,817.

Read the tank column carefully, because three of a Vengeance DH's seven axes do not measure what their names suggest — recorded on the 2026-08-28 Temple of Sethraliss record. `mitigation` has been saturated near 100 for every tank in the archive since the aura-parse fix, so 92 is not a distinguishing result. `utility` prices only two buttons and excludes sigils entirely, so 83 is close to meaningless. `damage` had its band cut on a corpus that was almost entirely this tank in +2 to +9 keys.

One correction to that older note while we are here: its arithmetic is now superseded. A raw share of 0.665 scored 84.6 on this record, where the old linear 0.36–0.80 band would have returned 69.3. Scores pass through the calibration curve now. Either way, do not read damage 85 as an instruction to do more damage — the archive control across 46 keys shows three non-DK tanks all landing near a 0.55 share while two Blood DKs sit twenty points clear, which points at spec rather than at effort.

What is left and is real for this tank: survival, mechanics, activity and interrupts. All four are at or near their ceiling. There is nothing in his execution to change. The thing to change is what he pulls.

B (76). Throughput 76 means 77% of the damage that actually needed an outside heal got covered — 141,599,923 of it, with 44% of the party's intake either self-covered or overkill. Response 96 means output rose sharply in the party's worst windows, putting back 51% of the 97,352,368 that landed in them.

A first pass suggested Aura Mastery was being held too long — presses at 1:01, 4:46, 8:26, 11:48 and 14:52 against a 180-second cooldown. Tested against the log rather than asserted, that is wrong. Damage landing in the 8 seconds each press covers totals 24,850,825 as actually played, against 14,894,747 for a strict 180-second cadence, which would have fired the last one at 16:01 into nothing at all. The holding was a good read of the fight and should be kept.

Lay on Hands is a real gap, but not a discipline one. Two casts, 0:33 and 8:45, a 492-second gap — which rules out a 600-second cooldown and puts it at 420. It was therefore available from 15:45, and the next deaths were at 17:49. There is no failed Lay on Hands cast anywhere in the log, so nothing refused it; it was never pressed. Forbearance was on nobody at 17:49.

The reason is the interface. The TollPips panic entry prompts Lay on Hands at 40% health. At 17:48 Shadow Priest read 41%, then 25%, then 11%, then dead — about a second and a half of warning, and by that addon's own constraints a health-threshold cue cannot carry sound. Toxic Aura had been on all five players since 17:37. The warning existed twelve seconds before the cue fired.

The tables cannot know when they are wrong for a particular run. Three cases here.

`dispel 69` is close to being the wrong axis. It scores 12 cleanses against the 19 a single dispeller could physically reach, which is arithmetically fair. But it frames the problem as a dispel shortfall and it was not one. No dispel number available to one healer answers five applications in four seconds against an eight-second Cleanse. The fix was never more cleansing.

The tank's `mitigation`, `utility` and `damage` are covered above and should not be read as performance signals.

And the run is `abandoned`. Every rate on this page is measured over 18:22 that stopped mid-key on a boss the group did not kill. It cannot be compared with a timed key, and it is not evidence about how this group performs over a full run.

Worth keeping for the next Voidscar attempt, in rough order of value.

1. The Agitated Voidscythe is the single highest-value thing to route around. It is an ordinary pull mob, not a summon, so it can be avoided or taken alone. Four logs above +19 contain none.

2. If one is unavoidable, never hold two. The entire difference between a three-target wave and a five-target wave is whether a second one is alive, and the five-target waves are what killed the key.

3. Corrosive Essence needs a named kicker, ahead of Lava Bolt. Eleven casts went out unstopped while 19 interrupts went to two other spells.

4. Pull 3 wants breaking in half. 25 enemies over 279 seconds is the slowest stretch of the key per enemy killed, it took a quarter of the run's damage, and it cost three deaths.

5. The opener deserves a separate look. It leaked 18 of the run's 21 unstopped casts and dropped the tank to 11.9%, which spent the healer's whole emergency kit inside the first minute and left the next five minutes without it.

6. Routing time was 57.5 seconds out of 18:22. There is no speed to be found in the path. Everything available is in pull composition and interrupt assignment.