Hi Andrés,
Please note that if you work according to the document, when restoring the replication you will lose the information added to the secondary database during the time the primary was inactive. When the secondary and processing nodes was active, audit records were written in the secondary database. When replication is restored, the primary database will be cloned to the secondary database and will wipe out the data added on the secondary while it was the only one running.
Our official documentation recommends another way to perform upgrade on cluster, but does not suggest a procedure to do it without downtime. I will send a request through our internal channels to amend it in a way that will also instruct how to do it without downtime, but if you could submit an 'idea' to the community to add it to the documentation as well, it would assist us prioritising it. I will be the first one to vote for that idea.
Thank you for bringing that up,
—Samuel
Vote for my feature suggestions