If the TD were to balance the tables as players bought-in, then the first 10 players to buy-in would be at table 1, the next 10 players would be at table 2, and the final 10 players would be at table 3. Not very random.
Actually, what would more likely happen is the first 10 players to buy-in would be at table 1. But when the 11th person buys-in, 4 players would be moved from table 1 to table 2, to balance the 11 players at 2 tables. When the 12th player buys-in, he would be placed at table 2, the 13th at table 1, the 14th at table 2, etc, until 2 tables are filled. When the 21st person buys-in, 2 players from one of the tables and 3 players from the other table would be moved to table 3, to balance all 3 tables (7, 7, and 6). Then repeat the process as the remaining players buy-in.
The point is, you'd have players seated, then moving, possibly twice, before the tournament even started. The TD takes a different approach: if you know how many players will be showing up (or approximately how many players), you can randomly seat them at buy-in. Because balancing isn't considered, the players are randomly seated and won't need to be moved before the tournament starts. If you don't know how many players will be showing up, you're probably better off NOT seating at buy-in, but waiting until buy-in is closed and then seating players.