# Reward split ignores slashed or outlier inputs

Divisors and reward splits that count skipped, slashed, or outlier participants as full weight

_Area: DeFi accounting and math bugs. Seen in 3 disclosed reports; latest 2026-05-18. Updated 2026-10-30._

## How it works
A reward or gas split assumes every participant in the divisor carries equal, valid weight. When that assumption breaks — a bond is slashed, a slot is an outlier, a ratio is intermediate — the divisor still counts the participant at full value while the numerator (or the actual paid set) does not. Value is silently redistributed: to the wrong recipients, or back to the payer.

## Where it shows up
- Division by a receiver count or replication factor where the paid set can shrink independently of that count.
- Skip/continue paths in a consume or distribution loop that do not adjust the divisor.
- Comments asserting "all bonds are equal" or similar, with no query of actual weights.
- Reward formulas using divisors or reciprocals where a spec describes linear scaling.
- Slashing, withdrawal, or partial-unstake logic that changes weights without touching the split.

## How to review
Identify the divisor and the set it is meant to represent. The invariant: the divisor must equal the sum of the actual weights being paid, and every skipped or reduced participant must be excluded from it. Check each skip path and each weight-changing event (slash, unstake) against the split. For published formulas, translate the spec into exact integer arithmetic and compare curves across the full input range, not just endpoints.

## Variants seen
- Divisor counts outlier slots that consume no gas, shrinking honest executors' share and refunding residue to the requestor [E2](https://bugbountylibrary.com/r/sherlock%3A2024-12-seda-protocol%3AM-3.md).
- Equal division by receiver count ignores slashed bonds, so a fully-slashed operator still earns a full share [E3](https://bugbountylibrary.com/r/c4%3A2026-01-olas%3AM-08.md).
- Divisor-based multiplier diverges from a documented linear reward spec, underpaying most stakers [E1](https://bugbountylibrary.com/r/c4%3A2023-03-mute%3AM-07.md).

## Impact and severity
Honest participants are underpaid and value leaks to the payer or to ineligible recipients [E2](https://bugbountylibrary.com/r/sherlock%3A2024-12-seda-protocol%3AM-3.md)[E3](https://bugbountylibrary.com/r/c4%3A2026-01-olas%3AM-08.md); nearly all stakers lose rewards versus spec [E1](https://bugbountylibrary.com/r/c4%3A2023-03-mute%3AM-07.md). Severity rises with the fraction of affected participants and when the leak is continuous rather than one-off.

## Why fixes fail
No fix evidence in these reports.

## Evidence
- E1: https://bugbountylibrary.com/r/c4%3A2023-03-mute%3AM-07.md
- E2: https://bugbountylibrary.com/r/sherlock%3A2024-12-seda-protocol%3AM-3.md
- E3: https://bugbountylibrary.com/r/c4%3A2026-01-olas%3AM-08.md