New issue by loreb on void-packages repository

https://github.com/void-linux/void-packages/issues/37382

Description:
<!-- Don't request update of package. We have a script for that. https://repo-default.voidlinux.org/void-updates/void-updates.txt. However, a quality pull request may help. -->
### System

* xuname:  
  Void 5.10.117_1 x86_64-musl GenuineIntel uptodate rFFF
* package:  
  nix-2.8.1_1

### Expected behavior
It works as usual

### Actual behavior
```
$ nix-channel --update
unpacking channels...
RT34
$ echo $?
162
$ strace nix-channel --update
(...)
vfork()
+++ killed by SIGRT_2 +++
RT34
```


### Steps to reproduce the behavior

Run the above on a musl machine.

To make it more interesting, if you stop the nix-daemon service,
log in as root, and run /etc/sv/nix-daemon/run by hand the behavior changes:
nix-daemon starts to spew many, many, *many* messages saying that it accepted a connection from $pid until it runs out of memory while keeping tens of thousands of nix-daemon processes alive (or zombies?); I have no idea if adding substantially more memory would help, but I know at least that this used to work before.

The bug happens the exact same way with the newly released 2.9.0 (I believe - I had to fiddle with the `--with-sandbox-shell` option to get it to compile for some reason).

`grep -R SIGRT` in the nix sources doesn't return any result, so I guess it's in some dependency (boost?)

Any help is greatly appreciated.