Cluster: Multi-opponent player_scope resolution stops after the first opponent

[matthewevans]

Rationale

A family of effects that should iterate over EACH opponent only resolve against the first. Shared root: the ResolvedAbility player_scope driver / per-opponent iteration loses subsequent opponents in 3+ player games. Herald of Hadar is code-confirmed (game/effects/roll_die.rs ~L92-100 drops player_scope when building the branch ResolvedAbility). Cluster end: per-opponent iteration is centralized and all members iterate all opponents.

Likely seam: crates/engine/src/game/effects/roll_die.rs (confirmed) + the ResolvedAbility player_scope driver in game/ (PlayerFilter::Opponent iteration)

Ordering

Start with the code-confirmed member (#1972) to localize the shared seam, then fan the fix across the remaining children.

Children

(attached as sub-issues)

[matthewevans]

Re-verification update (corrected pass): the "shared multi-opponent driver" hypothesis only partially holds.

• Herald of Hadar — When you roll the dice none of the opponents loose life and it makes the caster loose life instead. #2026 Herald of Hadar — CONFIRMED, but the bug is local to roll_die.rs:93: ResolvedAbility::new(...) drops player_scope: {Opponent} from the branch's AbilityDefinition, so each band's LoseLife/GainLife hits the controller instead of each opponent. This is a die-roll-branch-specific defect, not a shared scope driver.
• Extort (Crypt Ghast) is stopping after 1 opponent and not giving life — [[Crypt Ghast]] Extort trigger causes first opp… #1972 Extort (Crypt Ghast) and Aclazotz, Deepest Betrayal attack trigger stops after first opponent — Like the Extort trigger, [[Aclazotz, Deepest Bet… #1974 Aclazotz — downgraded to needs-repro: their player_scope resolution paths look correct on static inspection; the "stops after first opponent" symptom could not be confirmed from code and needs a 4-player repro.

Net: fix #2026 at the roll_die.rs seam on its own. Keep this tracker open as the multi-opponent-iteration watch, but #1972/#1974 may be a different root cause (or non-issues) pending repro — do not assume one fix covers all three.

[matthewevans]

All child issues are closed. Retrospective: multi-opponent player_scope resolution reports in this cluster have been resolved; closing the finite tracker.