The multiline codec ignoring blank lines side-effect was introduced in 0.1.4 on January 27 2015, to cope with inputs that supplied text with embedded newline characters - the upside, the text was split on newlines and the resulting array of lines was fed to multiline - the downside, text containing only one newline, when split gives an empty array, effectively ignoring the empty line.
It is a bit harder to say exactly which version of Logstash has this codec version.
I think it is reasonable to expect behavioural changes when making such a big step (from 1.5.X to 2.2.X)