Votes per machine

Regarding this quote from freepress.org/departments/display/19/2004/900

 "The reason that slightly fewer were processed per machine in high-D precincts could be because of differences in poll worker training and/or inefficiencies due to the large lines."

Another factor that would affect this is the actual number of working machines in each precinct.  Many reports from precincts with long lines stated that not all machines were working at all times.  How many reports of nonworking machines were received on election day and what was the response in each case?