Why do we think that three is the magic number when evaluating our ability to manage PostgreSQL relationships? The answer is reorganization.
If we ever need to utilize the DR environment, the entire primary system must eventually be rebuilt. Likewise, if a streaming replica is promoted, every server that once depended on the primary must switch to its stream instead. These actions take time and must be repeated at least twice as three or more servers are involved. Every time a command is manually invoked, there's a chance of a mistake.
Highly available servers do not have the luxury to withstand accidents. One misapplied stream change might spell the difference between platform errors, a system outage, or normal operation. ...