The known lifesteal setup comes from demo testing, not an official formula. Its reported strength was a flat heal on each enemy hit, which made broad attacks much more valuable than the same effect on a single target.
Reported interaction
A third-party demo review described a five-health-per-hit lifesteal buff that triggered for each target caught by a hammer swing. Against a group, one attack could therefore restore several instances of health. Official material does not confirm the number, source, or launch behavior.
Build for contact count
If the live effect still works per enemy hit, broad and controllable attacks should be tested before fast single-target weapons. Do not assume every damage pulse, environmental hit, or secondary effect triggers healing; verify with repeatable health changes.
Control before greed
Healing does not erase stagger, trap exposure, ranged attacks, or a lethal burst. Pair recovery with crowd control, armor, spacing, and a known retreat path. The purpose is to increase successful returns, not to justify carrying every haul into the next unknown room.
Furniture and altar interactions
Furniture can provide persistent combat bonuses and altars can radically alter playstyle. Check whether healing is a furniture effect, equipment property, altar result, or another system in the current build, because source determines availability and stacking.
Test protocol
Remove other healing variables, record starting health, hit one target, then several targets with the same attack. Repeat after changing one bonus. Test bosses separately.
A short table with build number and observations is more reliable than a single dramatic clip.
Status at launch
Exact lifesteal terminology, acquisition, value, cap, stacking, multi-hit behavior, and boss interaction are pending. This page deliberately avoids presenting the demo's reported five-health value as a guaranteed current statistic.
Evidence status
0 first-party developer updates currently map to this guide. Version-specific demo or third-party observations are explicitly labeled and are not treated as current official statistics. Unknown numeric values remain unclaimed until they can be verified in the current build.