Very roughly speaking, "defeasiblity" is about rules being able to prevent other rules from firing, or canceling the conclusions other rules.

It is used to manage "exceptions" or "(counter)arguments", where different rules lead to conflicting conclusions and a resolution strategy has to be applied, depending on some relative ordering (strength, salience, temporal, weight, etc..).

While an ad-hoc implementation is possible combining salience, appropriate insertions/retractions and the rule-based agenda, a more declarative approach should be provided to the user.

This lack of a feature is currently a blocker for the adoption of Drools in several medical and legal communities, just to cite some.