While I do not have access to AS 9120B, I presume due to its kinship to the other standards that usual calibration best practices apply to monitoring and measuring equipment. These do not only deal with what is in essence adjusting/confirmation for the coming period.
(Re-)calibration is also intended to confirm your presumption that the device was within acceptable deviation in the past period.
If it is found to not be within calibration, actions must be taken to assess the impact on past measurements, which might extend to field recalls in the worst case.
That calibration would also be executed when the tool is (potentially) damaged in a way that affects its function or reliability (e.g. because of a fall), or has experienced conditions for which it was not designed/specified that might influence outcomes after that fact (e.g. environmental excursions), thus not only when the period expires.
Lastly it should occur when terminating the use of the device, even when this is within the period that it was presumed to be in calibration.
Each of these leads to a possibility of change in or substantiation to keep its status identification.
Now I know that ISO 9001:2015 or :2008 requirements don't fully match to the above practices if you have a smooth talker on your hands, but it isn't about the requirement (often happy path based), but the value you gain from the principles they require. No-one can deny the need to have a process to accomodate for the above two unhappy paths for when unfortunate events occur, and the value they bring for reliability in your products.
I don't even know whether someone ever highlighted the risk in leaving this expectation unrequired/implicit (as safeguarding against (unintended) adjustements, damage or deterioration is not as clear-cut same as determine confidence upon detection of "", and the initiating trigger can be interval or prior to use based, without risk based considerations justifying why you can choose one or the other).
The validity determination upon finding unfit equipment is weak in this sense as it does not require when to scan for unfitness. In some cases a tool will function, but simply not produce reliable results, and neither does it require the need to calibrate when terminating a tool prior to the expiration of its current validity period. The best practice mentioned above fills these holes.