You can leave it.

As long as you ran systemctl daemon-reload, you should be able to try sleeping without needing to reboot.

It might be due to https://github.com/systemd/systemd/issues/33083.

Try disabling user session freezing when sleeping:

sudo systemctl edit systemd-suspend.service

Add the following to the file:

[Service]
Environment="SYSTEMD_SLEEP_FREEZE_USER_SESSIONS=false"

Reload systemd:

sudo systemctl daemon-reload

After that, try sleeping and waking again.

What filesystem are you using on the external drive? If it is NTFS or FAT, they won’t store permissions on the filesystem, which would explain why the owner/group changes are not persistent. To fix that, you can set the uid/gid on mount in your fstab.

/dev/mapper/YOUR_DRIVE /path/to/mnt <fstype> rw,uid=<jellyfin_uid>,gid=<jellyfin_gid>,dmask=0002,fmask=0113

Are you using the default bridge? I have a similar setup (with Traefik instead of NPM), and for each compose file am using separate networks for the internet, proxy, and backend services.

services:
  some_service:
    ...
    networks:
      - frontend_network
      - proxy_network
      - backend_network
  backend_service:
    ...
    networks:
      - backend_network
networks:
  frontend_network:
    driver: "bridge"
  proxy_network:
    driver: "bridge"
    internal: true
  backend_network:
    driver: "bridge"
    internal: true

It’s been a while since I took statistics, but yes, I guess that is a binomial distribution. It does not influence the results in the way you are implying it does, though. The calculator does actually account for it (the Population Proportion input), and the sample size actually decreases the lower/higher your proportion is. My point was that a question like, “Do you watch anime weekly,” is not like a question like, “How many hours of anime do you watch in a week,” where you certainly couldn’t assume a normal distribution for the number of hours watched.

Normal distribution with regards to what? “Do you watch anime weekly” is a binary question. There really isn’t a distribution associated with that.

You don’t need a massive sample size for surveys to give meaningful information. Play around with this sample size calculator if you want to see what the margins of error are: https://www.calculator.net/sample-size-calculator.html?type=2&cl2=95&ss2=4000&pc2=5&ps2=500000000&x=Calculate

the timer has no idea if it was triggered during last boot. It only has the context of “this” boot, so it will do it right after a reboot and set a timer to start the service again after a week of uptime.

This is not correct. Persistent=true saves the last time the timer was run on disk. From the systemd.timer man page:

Takes a boolean argument. If true, the time when the service unit was last triggered is stored on disk. When the timer is activated, the service unit is triggered immediately if it would have been triggered at least once during the time when the timer was inactive.

OP needs to remove Requires=backup.service from the [Unit] section so it stops running it when it start the timer on boot.

You have the timer requiring backup.service, so it will run that service every time the timer starts on boot. Remove Requires=backup.service, and that will fix the issue.