Erlang Distribution Cookie Handshake Fails in Production Cluster

An Erlang application works as a single development node but fails to form a distributed cluster in production. The distribution handshake depends on a cookie value configured with -setcookie or erlang:set_cookie, and all production nodes appear to use the same string. The unresolved behavior is whether runtime version differences, OS security policies, or node naming conventions can invalidate an otherwise matching cookie and prevent handshake.

The goal is to identify which factor breaks cluster formation when cookie values look identical. Constraints include avoiding a full node restart in production and verifying connectivity without disrupting running nodes.

Which runtime or OS factor can invalidate a matching distribution cookie? What role does node naming play in cookie derivation between environments? Which checks isolate the handshake failure without restarting nodes?