As someone in the middle of destroying a monolith I hope the title is not true. Of course distributed-yet-still-deeply-coupled systems are possible they are at least harder to create.
For me though the most important thing is grokability. Our monolith is to a point literally no one on earth can understand the whole thing.
Even if the system is complex, the individual deployables being fully understood by some number of engineers is extremely valuable and drastically reduces search space for the cases where things don’t go as planned
Introducing boundaries between problem domains should drastically reduce cross-cutting concerns. It also makes issues easier to find.
Not that this isn't possible in a "well engineered" monolithic system, but design constraints are usually better than hoping for engineering discipline.
For me though the most important thing is grokability. Our monolith is to a point literally no one on earth can understand the whole thing.
Even if the system is complex, the individual deployables being fully understood by some number of engineers is extremely valuable and drastically reduces search space for the cases where things don’t go as planned