I'd say that's a very reasonable and very good idea. I've implemented similar ideas in many other systems. Batching in general always reduces the per-entry overhead. It's easy to prove this via a benchmark. You can even put all the batched items in a single transaction. And since you are batching, you essentially will always have one thread that's getting the batch and actually doing the write, perfectly matched with SQLite's limitation of one concurrent write at a time.
What you need to worry about is slightly higher complexity: (1) what happens when a single batched write doesn't complete within one second; (2) what is the size of queue you store events in memory and whether it is unbounded or not; (3) if it is unbounded are you confident that overloading the server won't cause it to be killed by OOM (queueing theory says when the arrival rate is too high the queue size becomes infinite so there must be another mechanism to push back), and if it is bounded are you comfortable with dropping entries; (4) if you do decide to drop entries from a bounded queue, which entries you drop; (5) for a bounded queue what its limit is. These are very necessary questions that arise in almost every system that needs queueing. Thinking about these questions not only help you in this instance, but also in many other future scenarios you may encounter.
What you need to worry about is slightly higher complexity: (1) what happens when a single batched write doesn't complete within one second; (2) what is the size of queue you store events in memory and whether it is unbounded or not; (3) if it is unbounded are you confident that overloading the server won't cause it to be killed by OOM (queueing theory says when the arrival rate is too high the queue size becomes infinite so there must be another mechanism to push back), and if it is bounded are you comfortable with dropping entries; (4) if you do decide to drop entries from a bounded queue, which entries you drop; (5) for a bounded queue what its limit is. These are very necessary questions that arise in almost every system that needs queueing. Thinking about these questions not only help you in this instance, but also in many other future scenarios you may encounter.