Namespaced pages enable disambiguation - as the example in the Help database demonstrates, unrelated pages that happen to have similar common names can be viewed with exactly the same visible name - in the example, its “Tree”.
Less publicized but just as useful, namespaced pages facilitate finding connections as well. With longer and more useful page names we can create mini-structures for areas of a databases, broadening the potential link search space for pages without cluttering viewing those pages.
For example, someone was using Roam for knowledge management where a subtopic of another topic emerged, in his case “Economy -> Capitalism”. Now by creating three pages as below (and toggling namespace visibility down to the minimum), we will have:
With toggling, [[economy/capitalism]] will appear as [[capitalism]] just in a different color, with the economy/ prefix hidden. This means that the unlinked references of [[economy]] WILL include mentions of [[economy/capitalism]], with [[capitalism]] in a different color. There are a lot of ways to use this, depending on what one is trying to achieve. Finally, different forms of names (abbreviations, nicknames, initials, numbered foot/endnote references to a page) can be made more explicitly connectable to the visible name by namespacing. There can be more than one degree of namespacing also. For example: [[New York City/New York/NYC]]. If you see an unlinked reference you want to connect - be careful doing it automatically. Auto-linking to a page name that is also a component of a namespace will insert page links with full [] into the page name of the full namespace - messy in some cases. While wonderful for finding connections, consider linking specific pages explicitly.