cPanel server optimization challenge

We've been providing Website Maintenance & cPanel Management Services for a while now, and lately, we've hit a recurring wall with server optimization. Despite our best efforts, certain cPanel environments, especially those hosting demanding CMS platforms like WordPress or Magento, experience intermittent, severe performance degradation under what should be manageable load. This manifests consistently as high I/O wait times and unexpected CPU contention, particularly frustrating given the resources allocated. We've gone through the standard playbook: experimenting extensively with Apache MPMs like event and prefork, meticulously tuning PHP-FPM pool settings (dynamic/ondemand, max_children), and even deploying LiteSpeed or Nginx as a reverse proxy for static content caching. Database optimization has also been a major focus, involving meticulous tuning of MariaDB/MySQL parameters such as innodb_buffer_pool_size, query_cache_size, tmp_table_size, and various other buffer settings. At the kernel and system level, we've applied numerous sysctl.conf adjustments for network buffers (e.g., net.core.somaxconn, net.ipv4.tcp_tw_reuse), file handle limits (fs.file-max), and rotated through I/O schedulers, trying noop, deadline, and CFQ. For resource management, we've extensively utilized CloudLinux LVEs for strict resource isolation per cPanel account, believing this would prevent most resource contention issues. For monitoring and diagnostics, we employ a suite of tools including htop, iotop, netdata, New Relic, and cPanel's native metrics to pinpoint bottlenecks, yet the deeper cause remains elusive. Despite all these comprehensive approaches, we still encounter scenarios where a sudden, moderate traffic spike on a single cPanel account can inexplicably ripple through the entire server, causing system-wide I/O wait spikes and CPU load averages to skyrocket, even when overall server resources appear to have significant headroom. The issue isn't consistently tied to a single application or a specific database query; it seems to be a more fundamental interaction between the kernel, cPanel services, and potentially disk I/O under specific, intermittent loads. Standard server optimization guides often focus heavily on application or database tuning, but our challenge with these managed services feels deeper, touching on kernel scheduling, cgroup interactions, or perhaps less common cPanel service configurations that aren't widely documented. Are there advanced, perhaps unconventional, kernel-level or cPanel-specific server optimization techniques, beyond the usual suspects, that you've found remarkably effective in mitigating intermittent I/O wait and CPU contention issues on high-density cPanel servers? We're particularly interested in insights into optimizing kernel scheduling for I/O, advanced cgroup configurations, or any obscure cPanel service settings that might significantly impact overall server stability and performance under stress.