Understanding irq_exit() and non-interrupt context

Kernel newbie here. Started reading the kernel source a few days back and noticed that irq_exit() has code wrapped in an !in_interrupt() check:

https://elixir.bootlin.com/linux/v5....softirq.c#L413

When would we end up calling irq_exit() in a context where in_interrupt() would return false?

Ok, since you got no answer:
I suppose it's just a safety measure to catch the truly unexpected and unplanned event where something "out there" has already gone wrong. But I'm not a kernel developer, so probably they know it better.