Possible optimizations I could think of:

It turns out that

PropertyList[{g, VertexList[g]}]

can be used as well, and is slightly faster. This form does not seem to be documented.

Another possible optimization is to check if a graph has any custom properties. If not, do no try to retrieve properties for each vertex/edge separately, and save time:

hasCustomProp[g_] := OptionValue[Options[g, Properties], Properties] =!= {}

standardVertexProperties = {
  VertexCoordinates,
  VertexShape, VertexShapeFunction, VertexSize, VertexStyle,
  VertexLabels, VertexLabelStyle,
  VertexWeight, VertexCapacity
};

ClearAll[vertexPropertyList]
vertexPropertyList[g_ /; GraphQ[g] && hasCustomProp[g]] := Union @@ PropertyList[{g, VertexList[g]}]
vertexPropertyList[g_ /; GraphQ[g]] := Intersection[PropertyList[g], standardVertexProperties]

This seems a bit risky because it relies on an explicit list of standard property names, and the assumption that PropertyList[g] will indeed returns these in all cases. Is this true? Can anyone find a counterexample?

Yet another possible optimization is to only try to retrieve custom properties for vertices/edges that have them:

Intersection[
 Keys@OptionValue[Options[g, Properties], Properties],
 VertexList[g]
]

then use this in PropertyValue[{g,...}]. The problem with this approach is that when all vertices/edges has a property, it will slow things down. It will speed things up only when some of them do not have it.

Is there anything else I can do to improve performance? Do you see any potential problems with the above approaches?