math/rand always uses a deterministic seed. The sequence of session IDs
was therefore always the same:
Uv38ByGCZU8WP18PmmIdcg
lWbHTRDYaB0NhtHpHgAWeQ
...
Multiple copies of this program would have had session ID collisions. I
don't know what the consequences of that would be.