D-lock "bypasses" RCDs without actually removing them from the circuit, by passing a large DC from neutral to earth, saturating the iron core of the RCD, so preventing the trip operating. (this is applied asa a slow ramp to avoid tripping when 'D-locking')
When doing this deliberately, for a 200mA AC test, quarter of an amp or so is enough to lock a 30mA RCD, but much more fault current can be tolerated providing the fault current is in the same direction as the locking current - indeed a one half cycle 25A ELFI test is possible through an RCD locked in this way. Otherwise for AC leakage peak to peak needs to be less than the locking current, otherwise the core comes out of saturation on the cycles where fault current and locking current oppose. If this happens, it trips.
Such 'never trip' boxes are jolly useful for stopping nuiscence tripping too, and effectively make a 30mA trip into a 30mA plus root 2 times locking current trip, but, like bypassing the RCD with paper clips, this is only really a very nasty temporary measure (albeit a handy one if it is not your RCD that keeps interrupting your power). Keep the duration of your testing with D-lock on as short as possible, particularly if the building is occupied.
Note also that the hall effect based RCDs cannot be Dlocked, just the 'normal' bi-filar current transformer type.