Has anyone experience to confirm/contradict the thinking above?
Yes, I'm trying to understand what redundancy you're trying to achieve; and as importantly what redundancy the original deployment was trying to achieve.
Then: what does this have to do with the SQL DB, for which replication was allegedly used, but which isn't discussed in your post?
Also, has anyone ever used MySQL where is has stored a huge text file in addition to the actual DB Table data?
I have to admit, I'm unduly exercised to know what's in that file. But just like I don't have a mustache, it's not important. What is probably important is what process created that file. The claim is made that it's a text file; in my world no claim like that would be entertained without some explanation; if it was a textfile then it's puzzling why examination of its contents hasn't yielded clues as to its origin. I don't care how big it is: are you going to keep it around? If so how are you going to justify that decision to privacy and security stakeholders without understanding what's in it? Why do you claim it was created by MySQL?
Where is the MySQL replication instance? Is that file, mayhaps, a replication log that got so out of control it can't be replayed to bring the new repl into sync? If so, that was a mistake. But was it the cause of the complaint that "replication always caused issues"? Or is that a network or hardware reliability issue?
In general, replication (of the contents of a DB) involves replaying, in order, the commits which altered the contents.
In my world ((service + backing store) * 2) where the service and the backing store are cross mirrored so that the service is primary on one instance while the backing store is primary on the other suggests that there has to be a plan for load shedding if they both need to fail over to one or the other instance; a third node suggests tie breaking for a cluster.
By the way, for single writer or (mostly) read only loads, MyISAM gets an undeserved bad reputation. I say this as somebody who worked with RDB before Oracle bought the Koda engine, and long before anyone ever thought of MySQL, and they in turn thought of InnoDB.