I've got the library I wrote for #SuperCollider to keep installations running.  It checks if the sever has died and reboots it and checks on your pbinds and all your threads and if it can't get going, it drops out of sclang entirely so a shell script can restart everything. It's too clever by half.

So if it reboots the server, it reloads all your synthdefs. And then I realised it needs to know about buses and groups. But groups have this hierarchy and what if we accidentally store them in a different order than they were created? So I came with up with an elaborate lovely way of saving and reconstructing group trees. Took up most of my day, but will surely save my bacon later.

Then I went to walk my dog and realised I could just let the user write their own action for when the server restarts, as they'd probably rather manage their own damn groups.

And this is why walking regularly is good.