Public services could all go down together if one server had a problem

Every public-facing service reached the network through a single, specific server. If that one server had any kind of problem, every public service went offline at once, with no automatic recovery until someone manually redirected traffic elsewhere.

Traffic now reaches the platform through a floating network address that automatically moves to a healthy server if the current one stops responding, removing this single point of failure. Verified with a live failover test: recovery took well under a second.