I'm fairly ignorant of how FastCGI works and only slightly less ignorant about Warp after reading a paper on it by Michael Snoyman, but concurrent connections is a concern. I haven't been able to test the app with a large dataset but the algorithm is something along the lines of O^2 complexity so some requests may take much longer than others. One of my concerns then is that the server should be able to handle concurrent connections well and not let any long-running requests starve/block the others. If FastCGI forks off one process per connection, that would likely be a lot of overhead for the simple case.