A couple of remarks for DCA.
The appears to be an increase in cache size around 10%.
Also there is a performance regression (around 5% for C# - a bit more for Java, but that could be variance as the queries are executed on a small project).

For C# DCA reports a minor discrepancy between the generated models.
AFAIK after the new API was introduced some results could be missing when using the "old" taint tracking configuration.
Manually inspecting the reported difference in summaries supports this. The .NET runtime models have not been updated since the new API was introduced and most of the models reported as a difference was generated before the introduction of the new API (they have disappeared, but now they are back again).
However, there are a couple of extra models being generated, but they look sound:

System.Reflection.Emit;EnumBuilder;false;get_AssemblyQualifiedName;();;Argument[this];ReturnValue;taint;generated
System.Reflection.Emit;EnumBuilder;false;get_FullName;();;Argument[this];ReturnValue;taint;generated
System.Reflection.Emit;TypeBuilder;false;ToString;();;Argument[this];ReturnValue;taint;generated
System.Reflection.Emit;TypeBuilder;false;get_AssemblyQualifiedName;();;Argument[this];ReturnValue;taint;generated
System.Reflection.Emit;TypeBuilder;false;get_FullName;();;Argument[this];ReturnValue;taint;generated

DCA looks similar to the first execution.
Merging.