LLVM faces several significant challenges despite its widespread use and large contributor base. One major issue is the lack of review capacity. While many people contribute code, fewer are available to review it, leading to delays and potentially poor changes being accepted. New contributors often struggle to find appropriate reviewers due to a system that places the onus on them to request reviews. Implementing a more structured PR assignment system, similar to what Rust uses, could alleviate this problem. Churn in the LLVM C++ API and IR is another concern. Frequent changes keep the project dynamic but create difficulties for users, especially those with tightly integrated systems like downstream backends. While LLVM’s development philosophy encourages upstream contributions, it can impose costs on users who must continuously adapt to changes. Build times are impacted by the project's size, exceeding 2.5 million lines of code. Compiling LLVM can be a headache on low-spec machines, and even experienced users face challenges when building with debug info. Improvements like pre-compiled headers and a default dylib build aim to reduce these burdens. The CI system, with over 200 buildbots, struggles with stability and flakiness. It's common for commits to trigger build failures, making it hard to identify real issues amid the noise. Although pre-merge testing has helped, the scale of contributions complicates maintaining a consistently green CI. Testing is rigorous for individual components but lacks comprehensive end-to-end coverage. The separate llvm-test-suite repo has limited tests, missing critical operations across different data types. Lastly, backend implementations vary widely, leading to inconsistencies in optimizations and fixes, often only applied to specific targets rather than uniformly across the board.