diff --git a/Research/Borland C++ decompilation.md b/Research/Borland C++ decompilation.md
index dd64ea9a..4483323e 100644
--- a/Research/Borland C++ decompilation.md	
+++ b/Research/Borland C++ decompilation.md	
@@ -429,10 +429,6 @@ foo_c(100, 200); // PUSH 200; PUSH 100
 
 ### `-O` (Optimize jumps)
 
-Also merges individual `ADD SP, imm8` or `POP CX` stack-clearing instructions
-after `__cdecl` function calls into a single one with their combined parameter
-size.
-
 Inhibited by:
 
 * identical variable declarations within more than one scope – the
@@ -459,6 +455,24 @@ Inhibited by:
 
 * inlined calls to empty functions
 
+`-O` also merges individual `ADD SP, imm8` or `POP CX` stack-clearing
+instructions after `__cdecl` function calls into a single one with their
+combined parameter size.
+
+* **[Bug:]** Any emitted call to `SCOPY@` will disable this feature of `-O` for
+  all generated code in a translation unit that follows the `SCOPY@` call.
+
+  This can explain why a function might seem impossible to decompile with the
+  wrong translation unit layout. If it
+
+  * *doesn't* contain the stack-clearing optimization,
+  * but *does* definitely contain optimized jumps,
+  * which couldn't be reproduced with the slight jump optimization provided by
+    `-O- -y`,
+
+  the translation unit is simply missing a `SCOPY@` before the function in
+  question.
+
 ### `-y` (Produce line number info)
 
 Provides its own kind of slight jump optimization if combined with `-O-`. Yes,
@@ -650,3 +664,11 @@ contains one of the following:
 
   **Workaround**: Not happening when compiling via TASM (`-B` on the command
   line, or `#pragma inline`).
+
+* Any emitted call to `SCOPY@` will disable the stack cleanup optimization
+  generated by [`-O`](#-o-optimize-jumps) for all generated code in a
+  translation unit that follows the `SCOPY@` call.
+
+----
+
+[Bug:]: #compiler-bugs