In latest version (75422af ) of class MultiSCMChangeLogParser, line #73-76:

while(length > 0 && Character.isWhitespace(data[startIndex])) {
    startIndex += 1;
    length -= 1;
}

This removes all white spaces (including newline, etc.) regardless if the data is the start of the whole section.
This randomly cause problem, depends on SAX: When it pass everything in one data[], then it's fine. But sometimes, a section can be devided into several small pieces, and the white spaces in the middle can be incorrectly removed.
For example, the following data from git log:

commit b914376ca8a58c146636a84de9fa6d5acb8e020f
tree 43712b02868755704390e4f51ab480a966980874
parent e94e7abe6bd1df3c02e862107c3ae676f450ed2d

…..

Will be parsed to GitChangeLogParser as:

commit b914376ca8a58c146636a84de9fa6d5acb8e020ftree 43712b02868755704390e4f51ab480a966980874parent e94e7abe6bd1df3c02e862107c3ae676f450ed2d
…..

And it can not be parsed correctly.

I attached a snapshot from Jenkins, when it happens, the "changes" looks like the snapshot.

BTW, the snapshot is from v0.3 in our production system, but as far as I see, the latest version also have the same problem.