It could be a lamination issue - do you have it mounted so that the iron core laminations do not come into contact with something metallic - mounting it on a steel sheet might cause problems. Any conducting material touching/shorting the laminates could have this effect - think mounting brackets, fasteners thru holes in laminates etc..
Transformers are laminated to prevent large eddy currents circulating in the core - if the core was unlaminated solid iron/steel, it would be like one big shorted turn so, laminates are used for this type of power transformer and the laminates are electrically insulated from each other.
Any sign of rust on the edges of the laminates - any physical damage visible?
If in fact the transformer includes a bridge rectifier and smoothing capacitor then maybe the capacitor is beginning to fail - this could be the problem too. But it's still more likely to be an intermittent shorting turn on the primary winding as Spehro alluded to. A shorting turn (and it could be bridging many turns by the way) has the effect of changing the turns ratio so that when the short occurs, the output voltage will rise and the input current will also rise.
It's unlikely to be a secondary turn shorting because this will cause the output voltage to fall slightly and maybe also reduce the input current slightly (load dependent).