FM1 — Single Point of Failure
A load balancer that is not itself redundant becomes the single point of failure it was meant to prevent. The failure mode: the load balancer crashes or becomes unresponsive; all traffic to all servers behind it stops. The symptom is a total outage — all services appear down simultaneously — with no obvious cause in application logs because the requests never reach the servers.
FM6 — Hotspotting
IP-hash routing creates hotspots when client traffic is not uniformly distributed. If 30% of traffic comes from a corporate proxy with one IP, 30% of all requests go to one server. That server is overloaded; others are underloaded. The load balancer is routing evenly by IP hash, but the effective load distribution is highly skewed.