Imagine this is your original graph (not including the red node/edges), and A, B, C are your multiple sources. We can create an additional node X and the red edges shown, then do a standard BFS starting from X. Finally, all the distances will be 1 more than they should be, because you had the extra hop from X to the real source at the beginning of each path.

• Instead of pushing all the real sources into the queue, you add a fake source, link it to all the real sources, and just push that single fake source into the queue.
• If you do the BFS, it will take the fake source out of the queue first, and then it will push all the real sources into the queue right after (because you link the real sources directly to the fake source, they will be the closest to the fake source, which means they should be in the queue after the fake source is visited).
• Now, all the real sources are inside the queue, as expected. The side effect is that the calculated distances are all +1 compared to the actual distances, that's because the fake source and the fake edges increase the distance. So, subtract 1 from all the distances, and you are good to go.