Test Suite Optimization Using Redundancy Detection Coverage Analysis and Failure Prediction

Abstract

Purpose: Regression suites grow by accretion, and permanent test suite reduction promises lower execution cost, yet deletions
are rarely audited for the fault-detection capability they give up. This review asks when coverage-based, behaviour-based
and search-based reduction preserves that capability, and what evidence a permanent deletion should carry.
Method: A critical integrative review of software testing research, standards and adjacent measurement literature published
up to 2025, appraised for the validity of redundancy criteria, fault-seeding practice and version horizon.
Findings: Coverage equivalence is a weak proxy for behavioural equivalence. Greedy and exact set-cover solutions preserve
branch coverage while shedding tests that kill distinct mutants or expose distinct real faults. Similarity-based and multiobjective
methods, notably NSGA-II with mutation or fault-history objectives, retain more capability but depend on the
quality of the kill matrix and of failure history, which under-represents new and stable code.
Contribution: The article formalises reduction as constrained set cover with explicit loss accounting, proposes the Capability-
Preserving Suite Reduction (CPSR) framework with retention guards and a shadow-execution audit, and states six falsifiable
propositions with evaluation designs.
Implications: Deletion should be a reversible, audited decision: teams should report fault-detection loss on later versions
alongside wall-clock savings, and retain uniquely diagnostic tests even when their coverage is subsumed.