Load Balancing Software Resources

I've been researching load balancing options and wanted to open up a discussion on which solutions have real-time health checks and automatic failover that actually work reliably when a server goes down during peak traffic. The thing I keep running into is the gap between "supports health checks" in the feature list and what actually happens when a backend goes down during a traffic spike. Testing failover behavior in a controlled lab is easy. What's harder to find is honest feedback on how quickly traffic gets rerouted when a server drops without warning and everything is already under pressure.

Here's what I've been comparing:

1. HAProxy is the most cited option here, known for granular health checks, automatic removal of unresponsive servers, and zero-downtime reloads. Has anyone tuned the health check interval below default and seen a real difference in failover speed?
2. F5 NGINX is often picked for high-concurrency environments, with reviews citing 1M+ concurrent users across pods and built-in SSL offloading.
3. Progress Kemp LoadMaster stands out for its built-in monitoring dashboard that shows virtual service health and connection volumes at a glance, and is described as virtually maintenance-free once configured.
4. AWS Elastic Load Balancing covers health checks and graceful draining natively with Auto Scaling built in. Reviewers flag that debugging routing decisions without extra tooling is harder than it should be. Has anyone found a clean way to get that visibility without adding a separate observability layer?
5. Cloudflare Application Security and Performance sits at the edge and absorbs traffic spikes before they reach origin servers, which changes the failover model entirely.

Has anyone actually tested one of these during a real spike, not just a load test? Curious which held up and which surprised you.

I've been researching load balancing options and wanted to open up a discussion on which solutions have real-time health checks and automatic failover that actually work reliably when a server goes down during peak traffic. The thing I keep running into is the gap between "supports health checks" in the feature list and what actually happens when a backend goes down during a traffic spike. Testing failover behavior in a controlled lab is easy. What's harder to find is honest feedback on how quickly traffic gets rerouted when a server drops without warning and everything is already under pressure.

Here's what I've been comparing:

1. HAProxy is the most cited option here, known for granular health checks, automatic removal of unresponsive servers, and zero-downtime reloads. Has anyone tuned the health check interval below default and seen a real difference in failover speed?
2. F5 NGINX is often picked for high-concurrency environments, with reviews citing 1M+ concurrent users across pods and built-in SSL offloading.
3. Progress Kemp LoadMaster stands out for its built-in monitoring dashboard that shows virtual service health and connection volumes at a glance, and is described as virtually maintenance-free once configured.
4. AWS Elastic Load Balancing covers health checks and graceful draining natively with Auto Scaling built in. Reviewers flag that debugging routing decisions without extra tooling is harder than it should be. Has anyone found a clean way to get that visibility without adding a separate observability layer?
5. Cloudflare Application Security and Performance sits at the edge and absorbs traffic spikes before they reach origin servers, which changes the failover model entirely.

Has anyone actually tested one of these during a real spike, not just a load test? Curious which held up and which surprised you.

How good is Cloudflare DNS?