There are a number of possibilities. For example, the editor or the editor's assistant might have reset the number of referees needed, maybe to correct an initial error, or just incidentally by playing around with a different setting. It could be that the reviews were contradictory or that a review was not acceptable and needed to be replaced. It could have been a software glitch.
It could have been a change in associate editor.
Waiting for the result of a submission can be trying as much can depend on the acceptance or non-acceptance of a paper, for instance, if it is a condition for a degree. While the editing software tries to help by giving you information about the current state of your manuscript, it is often best for yourself to avoid checking it. The system is opaque and you can do nothing about the outcome. Use this information only if a decision is overdue. At least that what I learned to do myself in order to avoid fretting.
You could of course ask the editor, but I do not see what good can come out of it. The editor might not know because their view of the processing software is different and the editor is usually not the only one having access to it.