@Miroslav Cibulka It really is about your tolerance for downtime. Keep in mind that each additional PCE you join to the Supercluster will take a little longer to complete its join process, as it establishes replication with every other peer Supercluster member and replicates data from those PCEs to its local database.
Other than that, adding more than one member PCE during the same maintenance window while the entire Supercluster is in runlevel 2 should not be an issue, and in fact is the documented procedure in the PCE Supercluster Deployment Guide. Please note, however, that the actual process of joining member PCEs cannot be done in parallel - each PCE must be successfully joined one at a time until all new members that you wish to add are joined.