I've seen bearing failure in "real time" which caused the shaft to abruptly stop at 4000 rpm during run-down. The so called "haystack" on the spectrum became very prominent prior to failure. We were relying only on spectrum magnitudes and no enveloping was being done. Another time when I saw an ISB failure, a few days prior to the event, the high rotor vibration dropped abruptly and the low rotor vibration spiked suddenly on the strip chart at the same instant.