I don't like it, because a partially-wicked (
from both sides) via will trap gas and can create voiding or blowouts. Maybe it's less an issue here than for LGA and BGA: if that's a gull-wing style joint, the fillet probably closes up pretty quickly once a bubble goes past, and bubbles may have an easier time getting out the side fillet than they would a blind pad.
You're most likely doing it by reflowing one side, then the other, in which case the vias are plugged from one side, and then for 2nd reflow, you have the geometry of a one-side-tented via that traps flux/gas.
It might be viable to fill them with solder; this won't be easy to ensure with a lead-free process I think, but using larger diameter (0.4-0.5mm say?), and leaded solder, with suitable flux (if it's ENIG, pretty much any will do, but suffice it to say, a combination between solder and plating that gives good spreading), would give the best chance for it. A little excess paste may be helpful, but mind how it spreads out on heating, and it may be trading off with leaving solder beads on the surface.
Could also go over it by hand before assembly and fill and inspect, or even, like, shove bits of solder wire into the holes just to be absolutely sure it's pre-loaded, but, between the large labor input, and the potentially uneven surface before pasting, that's probably worse overall.
I did a design something like this some years back, it was for a gas-tight (not necessarily hermetic) joint between two grid-style connectors; I opted for blind vias, I think, and had enough room not to need filling or anything (dogbone traces). One thing that stands out in my memory, it was a, Samtec something connector I think, that suggested N2 or other inert atmosphere for reflow... fancy.
![Shocked :o](https://www.eevblog.com/forum/Smileys/default/shocked.gif)
Tim