What's the performance

A brief look through the codebase reveals that here we are using a lot of high-level commands to do the matching.

Org mode on the other hand constructs regular expressions and does the searching in very clever order to maximize performance (for example not even bothering with tags when doing full-text search since that has much higher hit-probabilty, doing the tag-match afterwards is cheap and fast post-process).

Did you do some benchmarks? My agenda spans hundreds of thousands of lines and so a slowdown even just 100% will be quite noticeable.