First I'd be leaving that one probe on the pulse train,
and start looking for other things in the neighborhood
which might cause it. That would be phase-synchronous
and preceding.
There might be a little bit of switching voltage coupling
that just can't be gotten rid of (close-in, radiated or
conductive). You might, if you find the roughly right
frequency signature somewhere, try things like disabling
its likely source and see if it disappears from the 'scope.
This is old school board level bench debug, have at it.