As Chrysalis said, it is likely down to the way the modems re-synced after the power outage. Your modem, like all others, will have measured the noise environment before any other modems were synchronised. They'd see a high signal, low noise environment devoid of crosstalk.
( A downside of giving everyone the same brand/firmware of modem - nearly - they all restart the sync process with identical timing)
Having detected low noise, the modems would probably all sync at an unnaturally high speed at first, but then they'd all be subject to high crosstalk noise immediately. Some would survive, like yours. Some would lose sync, and re-sync at a more appropriate speed.
Those who survive would now be operating at a low SNRM level, and be incurring high error rates (an SNRM of 2.5dB with an ES level of 148 in 47 minutes qualify as that).
The high error rate would have 2 impacts:
- in ongoing use, you'd find a small amount of packet loss. Something measuring speed tests might well encounter packet loss, which would cause TCP to regulate speed, so slower transfer speeds get measured.
- longer term, DLM intervention might be triggered overnight.
The DLM intervention is not a given, as the power loss event might well be visible amongst the stats to allow it to be recognised as a
wide area event.