since $t > s_i$ for every $i$, we can conclude that before the next process enters the system, the first process surely completes its execution (whatever scheduling algorithm we use, it will always be true).
So, at any instant of time there will only be one process in the system and before it completes execution no other process will enter the system, as $t > s_i$.
So, we don't need any kind of preemption. Preemption will only increase the complexity. At any time, since there will only be one process, the process will preempt to itself again and again until it finishes its execution, which will surely will reduce the efficiency .
Therefore I think the answer should be FCFS.