One recent example of spam edits has identified the user agent
http://code.google.com/appengine; appid: s~bagswish
in recent usage. It is reasonable to expect that such and similar types of abuse is going to continue to occur, and where it is occurring then it would seem an easy and convenient means to block such user-agent components after they have been identified by checkuser process
Would be useful to have variable_name user-agent, and variable values, with regex that could be applied.
Naturally there are some dangers, however, with the clear inability to be able to checkuser for a specific user-agent, and with such spammers having the ability to operate across so many wikis, the ability to prevent is required. I envisage that at WMF this would be a global filter added by stewards only as required.