'Thick' desktop apps often include a 'feedback' menu item, grouped under the 'help' or 'about' menu. I think that is about as close to a standard as you will find in that respect.
In another respect, if you have a new or troublesome feature, adding some method of directly launching that feedback interface might make sense. As it matures, remove it to clean up and further simplify the interface.
By embedding the activator for the feedback interface on each feature of interest or concern, you can do things like pre-populating the feedback form by including specific contextual information in a bit of a hidden state if it helps understand what the user was doing. Make the user aware each time you are gathering more information than they explicitly provide in the feedback form, perhaps even including a summary of that information. They will not like to find out later they were giving information seemingly without "consent."
There is a risk that by including this feedback activation at several points within the guts of the application that the user will activate the feedback interface for features that they want to give information or raise concerns about which do not directly correspond with the feature the interface was launched from. They may do this because the feedback activator they launched from was more convenient, most familiar with (or remember seeing the activation control for feedback in one spot but not others), or because they may not realize you populate the form with context information to a hidden or subdued area. With this in mind, be aware that the context information you pull in may not be valid or related in any way to the problem or concern or feedback at hand. It is only valid when the feedback interface is activated from the feature of concern.