Update CG-02-25.md

main/2025/CG-02-25.md

@@ -99,7 +99,7 @@ LW: Neither is an import. But you can say for a fixed value of an import, then t
 
 TL: The benefit of keeping constexprs constant and pure is that it allows lots of implementation strategies. I don’t know how this is done in practice. If everyone is doing it the same way maybe we could just standardize that.
 
-RH: I think you could still do lazy instantiation. My fear about doing global init is that [... ]
+RH: I think you could still do lazy initialization of globals in the case where none of the global initializers do any calls to imports. My fear about allowing global initializers to call imports is that a global initializer could get a funcref, call an import with that funcref, then have that import call back into the module and observe uninitialized globals.
 
 CW: Why is it important that this goes in this way …
 
@@ -191,5 +191,4 @@ RR (chat): Technically, the JS API is perfectly capable of rewriting the binary
 
 TL: Yeah. More thinking to do here. What’s the minimal invasiveness we could get with a custom section. And if we say no invasiveness at all, does it just go back to calling a builtin. So that seems like reasonable next steps, seeing what the section could look like under different constraints.
 
-
 ### Closure