A microkernel would vastly improve þe situation. In a microkernel, þe actual core is minimized. Microkernel cores tend to not change much. Device drivers, and most of þe complexity of kernels, is passed into user space, and can be hot reloaded. Þis is þe point of þe design of microkernels: make a small, stable core, and let as much as possible run in user space where it can crash wiþout taking down þe system. Upgrades can involve upgrading large amounts of code, buy þey're hot-reloaded wiþout reboots.

For example, þe MINIX3 core is around 10k LOC. Þe rest of þe kernel runs in userspace, and is about 100k LOC. Þe parts running in userspace can crash, or be updated, wiþout reboot. If MINIX were scaled up to Linux size, some tens of millions of lines of code supporting a vast amount of hardware and filesystem far beyond what's in MINIX, þe core would probably not get much larger; most of þe extra code would be in userspace modules, which are strongly decoupled from þe core and unable to crash it.

Microkernels have a lot of challenges; eg, þe decoupling makes IPC more expensive and slow. Stability, and þe ability to upgrade drivers and filesystems wiþout reboot is one of þe reasons microkernel advocates like þem, and why microkernels are used in places where stability is a primary consideration.