Legacy arrow critical clamp / Power enchantment compatibility investigation

This is a QA reproduction setup, NOT a gameplay fix. In-game validation has not yet been performed by the investigator.

Files:
- legacy-arrow-critical-clamp-repro-0.1.0.mcpack: data-only behavior pack; no RLCraft dependency or scripts. It supplies the stock legacy minecraft:arrow definition (format_version 1.26.0) retained in both examined installation archives.
- bow-damage-observer-0.1.0.mcpack: optional passive health observer (@minecraft/server 2.0.0); no entity overrides or damage-changing calls.
- native-comparison.txt: static evidence, package identities and limitations.

Suggested controlled reproduction:
1. Create separate clean test worlds on 1.26.30.5 and the latest stable 26.52 / Android build 1.26.52.3. Enable only the legacy-arrow reproduction pack. Do not enable RLCraft, the experimental damage compensation pack, other arrow overrides, or experiments.
2. Use normal difficulty, the same fixed position and an unarmored high-health target (e.g. a summoned iron golem) 8-10 blocks away. Use normal arrows and a bow with only Power V, with no Flame.
3. Fully charge every shot (hold at least two seconds). Test one hit at a time and measure target health loss. Use fresh targets as needed. Repeat at least 20 times per build.
4. Repeat with an unenchanted bow as a control. A lower damage result by itself is not proof; compare the same legacy arrow definition and conditions between the builds.
5. If using the optional observer, tag the player with bowprobe and the isolated target with bowprobe_target. Example commands: /tag @s add bowprobe ; /tag @e[type=iron_golem,c=1,r=15] add bowprobe_target . Compare complete hpPrev/hpNow/healthLoss records; eventDamage alone is not proof of actual health loss. Same-tick hits, healing or invalid entities after death can make the sample unsuitable.

Legacy player-arrow fields in this reproduction pack:
power=5, impact_damage.damage=1, power_multiplier=0.97, semi_random_diff_damage=true, min_critical_damage=9, max_critical_damage=10.

Conditional native-stage prediction from inspected code (not a runtime measurement):
1.26.30.5: critical min/max clamp first, then Power V multiplier 2.5 -> 22.5-25 at that native stage.
1.26.52.3: Power V first, then critical min/max clamp -> 9-10 at that native stage.
Armor, absorption, final rounding and script modifications can affect actual health loss.

The gameplay report that motivated the investigation involved RLCraft and was already observed in 26.40. The player's damage above 20 includes custom attack-level bonuses and is not a vanilla benchmark. This independent stock-legacy-definition setup is supplied so the engine compatibility question can be investigated without that third-party add-on.

Archive CRC, safe paths, JSON parsing and source/archive byte equality have been checked. Client import, runtime loading, and damage results have NOT been validated.
