But there is a problem with using freezing methods. The problem is that the system being imaged is static, literally frozen in place, and does not represent the dynamic state that exists in nature. It is like viewing a process in a series of snap shots set minutes or hours apart, and interpretation is required to infer the intermediate steps. This is particularly relevant for systems such as molecular interactions or assembly.
A recent paper in Science, and summarized in this article (RD mag nanotechnology) demonstrates a method to obtain high resolution images of dynamic processes in solution. The video below shows colloidal nanocrystal growth at high resolution in a specialized chamber composed of thin membrane layers of non-reactive graphene, which trap the liquid and prevents evaporation in the vacuum. The membrane is thin enough (1 atom thick) to not affect/have a minimal affect on the electron beam.
I think this is an amazing development and would certainly be interested to see if biological systems can be imaged in this manner. It is not clear how the beam affects the liquid sample and for biological specimens it may prove to be too destructive. However, overcoming the limitation of the vacuum is a significant step and this is certainly a technique to watch.