Since optical microscopy requires a shiny or lustrous surface, surface finishing methods like grinding, polishing and buffing normally used for metallic surfaces may not be effective for polymer composites. The abrasive grits on the emery paper or grinding wheel will leave behind scratch marks on the surface (polymers have poor abrasive resistance) or even leave behind stains which interfere with the surface morphology.

If one insists on optical microscopy, Clean the cured specimen surface with an appropriate solvent like acetone (depending on the resin system) and directly proceed with optical microscopy. 

In my opinion, Scanning electron microscopy (with sputtering) of the composite surface is better suited for microstructural studies.

That depends on the material and on what you want to see. If you want to observe the cross-section up to a certain magnification, it would be sufficient enough to grind the surface.

Since optical microscopy requires a shiny or lustrous surface, surface finishing methods like grinding, polishing and buffing normally used for metallic surfaces may not be effective for polymer composites. The abrasive grits on the emery paper or grinding wheel will leave behind scratch marks on the surface (polymers have poor abrasive resistance) or even leave behind stains which interfere with the surface morphology.

If one insists on optical microscopy, Clean the cured specimen surface with an appropriate solvent like acetone (depending on the resin system) and directly proceed with optical microscopy. 

In my opinion, Scanning electron microscopy (with sputtering) of the composite surface is better suited for microstructural studies.

Anand Pai has given the answer precisely. Clean with solvent like alcohol or acetone, if resin does not swell in contact with the solvent. If you have facility, apply few pulses of laser for ablation and test by microscopy.

Thermoplastics can be difficult as polishing can smear the surface, as identified by others already. However, polishing works very well for crosslinked polymers, as long as you are really careful about cascading down in grades of polishing paper/paste and clean thoroughly in between different grades. Ultrasonic bath recommended for such cleaning. 

Depends on what you want to see. In my case distribution of microspheres was the aim of main researches. I just poured composite on microscopic thin cover slip and end the reaction as usual (for thermoplastic cooling with furnace, for duroplaste - waiting few hours in dedicated conditions). I was watching composite through the glass. When I was working with nano-ceramics and polyurethane or/and phenolic resin, microtome with liquid nitrogen cooling chamber was the best tool for flat surface preparation (for AFM mostly but light microscopy too).