When a CLI only fails in one shell, what do you compare first?

Question about debugging CLI failures caused by environment drift.

A recurring debugging trap: the command works in one terminal and fails in another. Before changing application code, I usually compare binary path, runtime version, current directory, shell profile, project vs global executable, and config file path. The failure is often version drift or user-level config rather than the code being tested. For CLI failures, what do you compare first: binary path, runtime version, env vars, or config loading?