The failure review board report is an interesting read:
https://image.gsfc.nasa.gov/publication/doc...inal_Report.pdfQUOTE
The OFF condition of the SSPC is recoverable by cycling the command line to the device (which requires an OFF command followed by an ON command). For other than the Transponder SSPC, this could be performed via ground command. However, all ground commands are prevented from being received under the assumed scenario since the Transponder SSPC in an OFF state means that the Transponder itself is OFF (Receiver and Transmitter sections are OFF). Additionally, even if the command sequence could be sent from an on-board stored command load or macro within the SCU, it would be rejected by the PDU (by design). By design, the PDU is the only entity that can control the state of the Transponder SSPC and it will always try to ensure that it is ON, as discussed above. The only avenue identified that could cause the PDU to reset (OFF-ON sequence) the Transponder SSPC is a complete bus reset induced by a low voltage condition (<21 Vdc) which might occur during the next deep eclipse cycle in October 2007.
Or maybe it will finally undervolt sometime around 2017... this rings true. This computer was powered on, deaf, for 10 years, until the voltage finally degraded enough to trigger a reboot.
QUOTE
IMAGE’s orbital precession places the apogee in the Ecliptic for long durations at roughly 3 ½ year intervals. The resulting eclipse periods can exceed two hours in duration for which the spacecraft power system was never initially designed (due to the fact that IMAGE’s nominal two-year mission profile at launch would have never taken it through such a period). Similar long eclipse periods were handled during the extended part of the mission by payload pre-heating followed by intentional load shedding prior to the eclipse entry. This allowed for science instruments and bus components to retain temperatures above survival limits while also reducing battery draw down during the eclipse period. Due to the current situation this strategy cannot be implemented. However, analysis indicates that temperature limits to which observatory components were tested will not be exceeded during the next deep eclipse period. The fact that the spacecraft was not designed to handle deep eclipses is, in fact, what is vital to its possible revival. The next deep eclipse period will be in October 2007 with the longest period of darkness lasting ~160 minutes.
October 2007 April 2011 October 2014 April 2018... and the blog in the OP mentions an October 2016 precovery, so perhaps it was the 3rd eclipse that did it for a reboot.