I had an interesting experience with a 1/4" deep, keyhole, electron beam weld in 21-6-9 stainless steel. Can't recall the exact parameters, but basically it was found that slower travel speed (20 ipm) and a larger keyhole diameter (controlled by the beam focus) helped to minimize a servere problem with numerous large voids that were occuring at higher speed (50 ipm) and small beam diameter. My theory on this was that the large voids were produced by an unstable keyhole that periodically collapsed on itself and formed the voids. This was not a problem with 304 stainless at the high speed and small beam size, so I assume the the surface tension fluid flow of the 21-6-9 was different from the 304. Also, 21-6-9 has a high Nitrogen content (0.15 to 0.4 wt%), which probably contributed to N2 gas evolution from within the keyhole.
Also, there have been AWS articles written addressing high speed welding and problems with weld bead shape (humping) for various processes. I've seen humping with both GMAW and GTAW, induced by excessive speed.
Have you tried any experiments with your process?
In plasma keyhole welding, the trailing puddle will become an arrow pointing any from the arc just before you lose the keyhole. The keyhole method does not provide high travel speeds but does offer 100%penetration. The backbead of a plasma keyhole will disappear when you travel too fast. I always set the travel speed/current by the backbead.