M-x sly-disconnect crashes SBCL if connected via ssh tunnel

[nathanvy]

Hi. Thanks for all your work on sly, I really enjoy using it. This is a strange issue so I'll just jump right into it. To reproduce:

1. Install sbcl on Debian 11 Stable
2. $ sbcl
3. Invoke the slynk server:

   * (ql:quickload :slynk)
   To load "slynk":
   Load 1 ASDF system:
   asdf
   Install 1 Quicklisp release:
   sly
   ; Fetching #<URL "http://beta.quicklisp.org/archive/sly/2022-02-20/sly-20220220-git.tgz">
   ; 1792.02KB
   ==================================================
   1,835,032 bytes in 0.32 seconds (5549.49KB/sec)
   ; Loading "slynk"
   [package slynk-backend]...........................
   [package slynk-mop]...............................
   [package slynk-source-path-parser]................
   [package slynk-source-file-cache].................
   [package slynk-sbcl]..............................
   [package slynk-gray]..............................
   [package slynk-match].............................
   [package slynk-rpc]...............................
   [package slynk]...................................
   ..................................................
   [package slynk-api]...............................
   [package slynk-completion]........................
   [package slynk-completion-local-nicknames-test]...
   [package slynk-apropos].
   SLYNK's ASDF loader finished.(:SLYNK)
   * (slynk:create-server :dont-close t)
   ;; Slynk started at port: 4005.
   4005
   *

4. Establish an ssh tunnel to the server: $ ssh -N -L 4005:127.0.0.1:4005 ssh.example.com
5. Launch emacs and sly on local machine and immediately connect to the remote repl: M-x sly-connect RET RET
6. Immediately disconnect: M-x sly-disconnect

Expected result: REPL keeps running on remote machine

Actual result: SBCL on the remote machine crashes, with 300+ lines of nested errors in the stack trace. For legibility reasons I put it into a gist: https://gist.github.com/nathanvy/6a06f5362e9855b1cbf57b7a0d15a0f2

I have no clue how to proceed from here but it seems like sly-disconnect is causing the slynk server to choke when it tries to close a channel that's already closed.