There are indications that IDL generated sources can cause extreme slowdowns in compile times because source files may be scanned multiple times by multiple IDL generated sources. It's possible that we could precompile the most-used header files and greatly improve compile times overall. We currently do not have a way to tell which header files an IDL depends on from within SCons, though, which makes difficult an analysis of which header files may provide benefit if precompiled, or how we might be able to arrange some "common" header files for precompilation.

As a first step to improving build times, we should implement a way to read IDL files and determine a common set that we can precompile for maximum benefit. This needs only enable a proof-of-concept to see whether precompiled headers are of any benefit.