What are the specific use cases for this change?

I would think that in most cases this could just be done manually in edit mode.

In D4569#167179, @Sebastian Parborg (zeddb) wrote:

What are the specific use cases for this change?

I would think that in most cases this could just be done manually in edit mode.

Non-destructive, unlike doing it in edit mode.
Do things non-destructively as much as possible is always better.

In D4569#167185, @ThinkingPolygons (ThinkingPolygons) wrote:

In D4569#167179, @Sebastian Parborg (zeddb) wrote:

What are the specific use cases for this change?

I would think that in most cases this could just be done manually in edit mode.

Non-destructive, unlike doing it in edit mode.
Do things non-destructively as much as possible is always better.

That still doesn't explain the use case for this feature.

When would you like to triangulate only part of the mesh and in addition do so non destructively?

This could possibly be useful to achieve heterogeneous or pseudo-randomized triangulation for effects similar to what can be achieved with Tissue Addon.
It can also be used with different settings on different parts of a mesh for low poly style models

Being able to triangulate just part of the mesh could make the intentional creation of non-planar Ngons (as a modeling technique) a lot more practical. You can just triangulate the Ngons to make everything planar and leave the rest of the mesh alone.

In addition, partial triangulation could really help in non-destructively creating smooth inclines that curve (which require triangles if you want to minimize your polygon count).

Non destructive modeling has really taken off recently especially with addons such as Speedflow. From the discussion here, allowing the users to have more control is definitely useful.
Imagine this scenario, you have a mesh with multiple ngons (some with the same number of vertices), but you want to triangulate only some of ngons. To a certain extent, the Minimum Vertices parameter will help but it has its limitations and without stacking multiple triangulate modifiers it would be quite hard (if not impossible) to achieve the same effect as having a vertex group to mask the areas that should be triangulated.

I don't really see any issues with this.

@Brecht Van Lommel (brecht) any objections to merging this?

To me this seems like a poor fit. What this really needs is face groups, with vertex groups its unclear what the conditions are for a face to be triangulated.

Personally, I would wait with this kind of thing until we have modifier nodes and a better mechanism, unless someone can come up with a compelling example of why this is important.