Answer
A workable cadence for most IBM i shops is a full, timed DR test at least annually, with a lighter technical verification, such as confirming the recovery LPAR can be activated and reach current data, on a quarterly basis. Organizations under regulatory pressure, such as those in financial services, healthcare, or manufacturing with contractual continuity clauses, often need to test twice a year and produce documented evidence for auditors. The number matters less than the discipline of tying the schedule to a calendar trigger rather than waiting until things settle down, because things rarely settle down for long.
The bigger risk than an infrequent schedule is a stale one: a test plan that has not been updated even though the application portfolio, network topology, or staffing has changed materially since the last run. Buyers should ask how the DR software surfaces environment drift between tests, and whether it can flag when a new application, interface, or infrastructure change has not yet been incorporated into the tested recovery sequence. Software that makes each test faster to execute and easier to document, through automated status capture and repeatable checklists, is what actually gets testing cadence to stick instead of slipping every year.