One of the two UncaughtExceptionHandler implementations calls LOGGER.error(e). This only causes the toString of the exception to be logged, not the stacktrace. The affected implementation is the one obfuscated as e; the other implementation (f) calls LOGGER.error(t.getName(), e) and as such is not affected as it does log a message.

An easy would be to simply change it to LOGGER.error("Caught previously unhandled exception", e) instead of having that message on a separate line. Log4j should already put it on a new line, so this should be the same behavior.

An example of this in the wild is MC-127452 (I haven't evaluated that ticket in any other way; it just does have the message logged without a stacktrace).

As far as I can tell, the relevant code was added in 18w06a.