Apache Tomcat vs Caddy vs nginx

I was in a situation where I have to configure 40 RHEL servers 20 each for Apache HTTP Server and Tomcat server. My task was to

1. configure LVM with required logical volumes, format and mount for HTTP and Tomcat servers accordingly.
2. Install apache and tomcat.
3. Generate and apply selfsigned certs to http server.
4. Modify default ports on Tomcat to different ports.
5. Create users on RHEL for application support team.
6. other administrative tasks like, start, stop and restart HTTP and Tomcat services.

I have utilized the power of ansible for all these tasks, which made it easy and manageable.

I am diving into web development, both front and back end. I feel comfortable with administration, scripting and moderate coding in bash, Python and C++, but I am also a Windows fan (i love inner conflict). What are the votes on web servers? IIS is expensive and restrictive (has Windows adoption of open source changed this?) Apache has the history but seems to be at the root of most of my Infosec issues, and I know nothing about nginx (is it too new to rely on?). And no, I don't know what I want to do on the web explicitly, but hosting and data storage (both cloud and tape) are possibilities.
Ready, aim fire!

• Server rendered HTML output from PHP is being migrated to the client as Vue.js components, future plans to provide additional content, and other new miscellaneous features all result in a substantial increase of static files needing to be served from the server. NGINX has better performance than Apache for serving static content.
• The change to NGINX will require switching from PHP to PHP-FPM resulting in a distributed architecture with a higher complexity configuration, but this is outweighed by PHP-FPM being faster than PHP for processing requests.
• The NGINX + PHP-FPM setup now allows for horizontally scaling of resources rather vertically scaling the previously combined Apache + PHP resources.
• PHP shell tasks can now efficiently be decoupled from the application reducing main application footprint and allow for scaling of tasks on an individual basis.