The Fallacy of the Fully Connected Graph
When engineers and domain modelers sit before a blank canvas, an instinctive cognitive bias takes over: if two components exist within the same system, something must draw a line between them. This compulsion produces diagrams where everything touches everything, transforming what should be a sharp diagnostic tool into an indecipherable spiderweb. The presence of an edge implies direct coupling, synchronous invocation, or active governance. When you connect nodes that merely share a domain context or broad lifecycle scope, you destroy the distinction between immediate operational cohesion and passive co-existence.
Leaving nodes unconnected is an explicit visual statement, not an omission or unfinished draft. In clear visual semantics, empty space between modules signals loose coupling, independent lifecycles, and localized fault domains. Drawing an edge forces the reader to search for an interface protocol, an event dispatch, or a transactional boundary that simply does not exist in reality.
Core Diagnostic Rule: Spatial Placement Over Redundant Links
Contextual association, tenancy, and ownership are expressed more cleanly through visual grouping, spatial clustering, and dedicated containment boundaries than through line connectors that mimic execution flow.
The moment an arrow bridges two modules, every engineer reviewing the architecture must evaluate latency, failure propagation, and security handshakes. If the actual relationship is merely organizational proximity or shared infrastructure tenancy, the arrow creates phantom engineering requirements and muddies the mental model of the entire system.