Code generation and variable naming
For the longest time, in code, I never understood the purpose of ending identifiers with Id when they could be ended with ID. As a concrete example, turns out that in projects with code generation, proto models (in GRPC) are defined with snake case (entity_id). During code generation for HTTP endpoints (JSON models), entity_id is translated to entityId. This means there's at least one situation where being insistent on ID (as opposed to Id) makes things more difficult.
In short, when automation is involved, it's easier to rely on a straightforward heuristic to map between different casing of letters in identifiers.