If the web page test is valid and accurate, then the keyboard has no 6-key rollover. I made a bad assumption in the rollover definition. Henceforth I shall be better.
When I mash down large numbers of keys, I get seemingly random results from 0 to 6 keys. Zero appears more often than any other number.
When I type and hold, the timing is significant. Pausing will end the keypress reception.
Example
human press a,s,d,f,g,h
test result count 6
human press a,s,d, [pause], f,g,h
test result count 3
human press [numeric keypad ]
test result count 3 (maximum, all tests)
human press [anti ghosting keys q w e a s d {space} and {arrows}] any sequence
test result usually count 6, with pauses having almost no effect
human press [anti ghosting keys] + z x, any sequence
test result usually count 6, with pauses having almost no effect
This is not meant to represent scientific (or even consistent) testing at this point.
EDIT: clarified human touch parameters, and specified ghosting keys as shown on box.