LogbackMDCAdapter uses HashMap rather than LinkedHashMap and so the order of the key value pairs when logging MDC with %X cannot be predicted from one log line to another.  This makes log parsing harder than it has to be.

 Using LinkedHashMap would address this as the order of adding to MDC would be preserved.

See https://github.com/qos-ch/logback/blob/v_1.2.3/logback-classic/src/main/java/ch/qos/logback/classic/util/LogbackMDCAdapter.java

& https://github.com/qos-ch/logback/blob/v_1.1.11/logback-classic/src/main/java/ch/qos/logback/classic/util/LogbackMDCAdapter.java