Michal Koutný

The evaluated order (same as in `systemctl cat` output) is not correct, opposite the [intention](https://github.com/systemd/systemd/blob/59b13e07f229a7b8d1ef1565b64c4f019928443a/src/shared/dropin.c#L174). I can reproduce it with v254.5.

> How? The line (and the whole file) is about drop-ins, while the request was about the main config. You're right, I was wrong. There are not drop-ins competing. The...

@piperfw Can you please share `systemctl cat user-1001.slice` in the (unexpectedly) failing case?

> Changing this to user-.slice.d, the user-1001.slice.d overrides it as intended! Is that in line with what you expect? Yes. (Provided the drop-ins fall in the same `/etc`|`/usr`|`/run` category. IOW,...

The most useful statement in my experience is: check `systemctl cat $UNIT` to see what config files are applied (in what order) to the unit ;-)

> here does not to be any entry for user-1229790 That's weird. systemd should've created that slice. What is `systemctl status user-1229790.slice`? (Is it running?) Actually, I smell something fishy...

Also the [docs update](https://github.com/systemd/systemd/pull/36437#issuecomment-2725640227). I had an epiphany in the past why this may be needed (local DoS (all same timers exhausting the PID1 on one machine) vs distributed DoS...

OK, so I'd start with a picture like this: ``` i) t_{i-1} c_i t_i | |---- δ ----> v......^ ii) t_{i-1} c_i t_i | |---- δ ----> v......^ iii) t_{i-1}...

FWIW, I didn't drop the ball on this. I only got into the rabbit hole of proper testing timer elapse calculations (and more important stuff popped up too). The WIP...

@benjaminp What is the use case? The value of `memory.peak` is primarily useful to the actor that sets up `memory.max` (so write permissions should be set accordingly).