The key difference
Rectangle, Ellipse, Polygon and B-Spline are commonly used as masks. Their job is to describe an area: where Blur is applied, where a grade is visible, where a Merge is limited or where an alpha operation should act.
Shape nodes with an S prefix are different. SRectangle, SEllipse, SText, STransform and SMerge pass vector-shape data through a dedicated Shape pipeline. That stream is not a normal image and is not an ordinary mask stream.
What a regular mask does
A mask modifies the influence of another node. Connect an Ellipse to the mask input of Blur and only the ellipse region is blurred. Connect a Polygon to a Merge mask and the blend is constrained to that shape.
A mask therefore answers the question 'where should this operation happen?' It normally does not become the visible graphic by itself. If you want a visible filled or stroked shape as a design element, the Shape system is often the more appropriate toolset.
What the Shape system does
Shape nodes create vector objects that can be combined, transformed, styled and animated before rasterization. Several shapes can be assembled with Shape-specific operations and kept editable as vector elements throughout that branch.
This is useful for motion graphics, clean lines, repeated shapes, stylized typography and UI-style graphics where retaining a vector representation is more flexible than creating a raster image immediately.
Where the conversion happens
SRender is the important boundary. Before SRender, the branch is Shape data. After SRender, it is a regular 2D image that can be connected to Merge, Blur, Glow, Color Corrector and MediaOut.
Thinking in terms of that boundary makes a Flow easier to debug: if a regular image node refuses a Shape connection, the branch probably has not been rasterized yet.
When to choose each approach
Use regular masks when the main task is isolation or effect control: rotoscoping a region, limiting a grade, revealing part of a layer or constraining a transition. Use Shape nodes when the shape itself is part of the visible design.
A project can use both at once. A vector Shape branch can be rendered through SRender, merged into the shot and then constrained by an ordinary Polygon mask at a later stage.
Practical example: mask versus Shape
For a local correction, MediaIn can feed Color Corrector while Rectangle is connected only to the blue mask input. The Rectangle does not draw a visible layer; it simply limits where the correction is allowed to affect the image.
For visible motion graphics, SRectangle can create a plate, SText can create the label, SMerge can combine those vector elements and STransform can animate the block. SRender then converts the Shape result into an ordinary image that a regular Merge can place over MediaIn.
If those two approaches are mixed up, the Flow becomes confusing very quickly. A standard Rectangle is a control matte for another node, while SRectangle is actual vector content inside the Shape data system.