User blog comment:Rappy 4187/Technical Update: July 20, 2016/@comment-26154973-20160721213248/@comment-25683568-20160723084527

"if you were using regular templates with no lua, the results would be far worse"

And that is why we are using Lua...because it's suited for things regular templates aren't suited for...

The time and memory usage is variable, true. The variance isn't significant unless Wikia gets hit by enormous traffic so I don't see how that's important in this conversation though.

I know what the profiler is there for; I had a string formatting utility function eat up the entire alloted Lua running time due to an inefficiency from when I first started writing Lua for the wiki and that was very helpful. A shame there isn't a counterpart for memory usage. But the profiler can't tell inefficiencies from just a lot of operations in general, which is what I see in the July 20th NewPP limit report. Maybe I see different things in the profiler report than you do.

If the page is long by design, then I don't see how we can do anything to improve parser speed other than loading sections of the page via AJAX, which would throw mobile users out the window, or splitting into subpages, and then we are adding more clicks and page loads for users to get to what they wanted to see.