How are dependencies between files handled, on startup?

[gco-bmx]

I'm reading the codebase trying to understand and get a high level view of the steps, that occur for analysis

on start up it seems like all files are parsed in parallel,
design units added to root,
then analysis also starts in a parallelized way,
it doesn't seem to be ordered based on dependencies, since before analysis it appears to not be aware of what they are
so how does it work?

possible ways I guess could be
-the analysis of a file blocks when it tries to get something from another file that hasn't been analyzed yet,
-the dependence gets completely or partially analyzed, when it tries to be looked up

is it something like this or am I way off?