Mark,
As George indicated, most of your data is not versioned. It will not impact performance if you only have one or two feature classes registered as versioned; unless you have an enormous number of edits.
If you are only doing dozens of edits per week for each feature class, it should not be difficult to reconcile, post, delete versions, and compress weekly.
After doing your reconcile/post and deleting all versions after Default, I would suggest utilizing the ArcPy commands to complete your process. In simple form:
import arcpy
import time
from arcpy import env
workspace = r'Database Connections\SDE_CONNECTION_TO_DATABASE.sde'
env.workspace = workspace
database_name = str(workspace.split('\\')[1])
try:
print 'Blocking connections in %s' % database_name
arcpy.AcceptConnections(workspace, False)
except (arcpy.ExecuteError, arcpy.ExecuteWarning) as e:
print e
time.sleep(120)
try:
print 'Disconnecting all users from %s...' % database_name
arcpy.DisconnectUser(workspace, "ALL")
except (arcpy.ExecuteError, arcpy.ExecuteWarning) as e:
print e
try:
print 'Compressing database...'
arcpy.Compress_management(workspace)
except (arcpy.ExecuteError, arcpy.ExecuteWarning) as e:
print e
try:
print 'Accepting Connections from %s...' % database_name
arcpy.AcceptConnections(workspace, True)
except (arcpy.ExecuteError, arcpy.ExecuteWarning) as e:
print e
AcceptConnections—Help | ArcGIS for Desktop
DisconnectUser—Help | ArcGIS for Desktop
ListUsers—Help | ArcGIS for Desktop
You can of course also script your reconcile and post, but many organizations like to go through that process manually to verify/determine any conflicts.
In short, a few versioned feature classes will not adversely affect your databases performance; I would advise against creating a user schema geodatabase for this use case.
~Alex