I've asked similar question here: Is gossip protocol in Bitcoin perfect? But I didn't get an answer to my question really. I am ideating a consensus algorithm and I need definitive answer.
Can we take for granted that if in time period t there were 10,000 connected peers (no new peers connecting, no peers disconnecting), all with good internet connection and hardware, then if every of those 10,000 peers originated one transaction (or any kind of message) to his peers, can we take for absolutely granted that eventually every single one of those 10,000 connected peers have in his mempool all 10,000 transactions, including his own and those of 9,999 other peers and absolutely no transaction will be missing in mempool of any of those nodes (they will all have same set of data)? If yes, and would could be estimated time (seconds, minutes, hours)?
And now: things like double-spends, transaction correctness, peers connecting and disconnecting from network, etc. that happens in real cryptocurrency systems are not relevant here - this is a theoretical question and a theoretical situation.