When you view a truncated log file (Skipping ... Full Log) often the text will resume in the middle of a line. This is ugly enough but when the offset happens to be in the middle of an encoded ConsoleNote you get a confusing batch of Base64 garbage before the stream corrects itself.

Either console.jelly should employ a more sophisticated core API to find an offset which is at a line boundary (which might be inefficient); or AnnotatedLargeText.writeHtmlTo should skip any initial partial line, for example by using start - 1 and using a wrapper OutputStream which discards everything prior to the first newline character and thereafter mirrors its input; or ConsoleAnnotationOutputStream should have an option to do that.