As noted in the SPC discussion, destabilization ahead of a shortwave in the Ohio Valley spawned prolific downburst-producing storms, in an environment characterized by 20-30 kt of shear and strong low-level lapse rates.

...Ohio Valley into the Mid-Atlantic and Southern New England...
   On the southern fringe of upper troughing over eastern Canada and
   the Great Lakes, a weak/convectively augmented mid-level shortwave
   trough will progress northeastward across the OH Valley and central
   Appalachians through the day, eventually reaching the Mid-Atlantic
   tonight. Robust daytime heating of a modestly moist low-level
   airmass (surface dewpoints generally in the low to mid 60s) ahead of
   a surface cold front will aid in the development of moderate
   instability by early afternoon. Ongoing convection across OH/KY this
   morning may eventually strengthen as it encounters this
   destabilizing airmass, and additional thunderstorms are expected to
   develop along/near the cold front by mid afternoon from the central
   Appalachians into portions of the Mid-Atlantic and southern New
   England.

   Stronger flow aloft will tend to lag/remain north of the cold front,
   but modest enhancement to the west-southwesterly mid-level flow
   along/near the front should support around 20-30 kt of deep-layer
   shear. This will be sufficient for some updraft organization with
   mainly multicells expected, although marginal supercell structures
   may occur. Low-level lapse rates are expected to become quite steep
   through the day, evidenced by a general lack of clouds on recent
   visible satellite imagery from central/eastern WV into the
   Mid-Atlantic. This will likely aid efficient downward momentum
   transfer in convective downdrafts, with scattered severe/damaging
   winds possible as multiple clusters spread east-northeastward
   through the afternoon/evening. Isolated hail may also occur with the
   stronger cores. A Slight Risk for severe/damaging winds has been
   introduced from parts of WV to southern New England given increased
   confidence in multiple clusters traversing a narrow zone along/ahead
   of the front.

In the animation below, reports are valid for the next 45 minutes. Thus, the ProbSevere v4 predictions are matched up with what the model is trying to predict (any severe weather in the next 45 minutes and within 15 km). After playing the animation, you can see that PSv4 does an excellent job predicting the severe downbursts in northern Virginia, West Virginia, Maryland, Pennsylvania, and New Jersey. A few reports are missed (at the 25% threshold) in Virginia and New York.

At 1840Z, you can see the PSv4 probabilities matching up well with future downburst reports in PA, MD, and VA. The 25% contours cover all of the wind reports. There are no NWS severe thunderstorm warnings in this area at this time, indicating that the NWS was perhaps waiting a longer to gain confidence before issuing severe warnings.

There is also an area in central PA with over 50% probability, but no reports. There were wind reports later downstream of this region (see the animation above), so the PSv4 model was too early in this region.

At 20:16Z, the ≥ 50% contour has expanded in southeast Pennsylvania and northern Maryland. PSv4 is correctly predicting several clusters of wind reports in southeast PA with high probabilities. In Maryland, all five future wind reports (in the next 45 minutes) are covered by the cyan 25% probability contour, in a regions not currently under a severe thunderstorm warning.

A common trend I see, especially with linear or pseudo-linear features, is that ProbSevere v4 is out ahead of NWS warnings by a little bit (perhaps indicating the model trying to predict further out that most NWS forecasters), and NWS warnings also are slow to expire/reduce area on the backend (perhaps because forecasters don’t have a high incentive to reduce false alarm area or time). We see both of these features in New Jersey and Pennsylvania at 21:48Z.