The time when tighter coupling makes sense is when you are pushing the frontier of what is possible. In that situation attempting to enforce loose coupling pushes you to choose a way of modularizing before you understand the problem space well enough to pick a good modularization. Maintaining it also prevents you from using certain optimizations that you might come up with after you create your design.

This does not mean that people who are trying to accomplish the novel have free license to throw out the rule-book. To the contrary they are unlikely to successfully push those limits unless they try hard to make the complexities that they will encounter manageable. However they will also need to selectively violate normal principles just to make what they want to do possible.

However history shows that, when technology catches up to what they were trying to do, the people who created the integrated technology generally lose in the long run to latecomers with more modular technologies. Does this mean that they did something silly? Not at all! History suggests two other things as well. The first is that the recognition of the right way to modularize the problem is often only possible after someone has solved it badly, first. The second is that there is a lot of money to be made by being the first to be able to do something that people want to do.

For more on this, I highly recommend The Innovator's Solution. Be warned, it is a management book and not a technology book per se. However I think that what it has to say about which technologies are likely to replace others is useful for people who work with technology.