Not OP but I can answer with my own stats:

In just a week, With BTRFS compression (compress-force=zstd:3) & deduplication (via bees), media is at about 1GB (and I am subscribed to media-heavy communities like 196) and the postgres DB is at about 550MB (which is also currently shared with Matrix Dendrite)

At "idle" (as you can be while being connected to ActivityPub & Matrix), the immediate CPU and RAM usage breakdown per container is:

NAME        CPU %       MEM USAGE / LIMIT  MEM %       NET IO             BLOCK IO           PIDS        CPU TIME         AVG CPU %
pict-rs     0.20%       18.92MB / 4.005GB  0.47%       3.319GB / 1.105GB  17.58GB / 3.239GB  13          1h16m57.232828s  0.59%
crowdsec    1.39%       44.23MB / 4.005GB  1.10%       106.4MB / 23.46MB  25.53GB / 486.7MB  11          45m28.744419s    1.95%
caddy       0.63%       73.06MB / 4.005GB  1.82%       1.675GB / 1.977GB  3.322GB / 720MB    10          21m9.94572s      0.90%
dendrite    1.58%       197.7MB / 4.005GB  4.94%       912.8MB / 2.33GB   8.718GB / 4.761GB  12          53m26.302022s    1.43%
postgres    5.33%       82.51MB / 4.005GB  2.06%       56.22GB / 7.961GB  20.92GB / 295.7GB  23          8h20m28.078567s  2.86%
lemmy-ui    0.00%       48.71MB / 4.005GB  1.22%       3.491GB / 5.961GB  3.603GB / 5.267GB  12          31m35.884936s    0.24%
lemmy-be    2.82%       29.01MB / 4.005GB  0.72%       16.45GB / 57.85GB  7.966GB / 6.439GB  6           3h6m34.633508s   1.42%

Net IO you shouldn't really care about as that includes inter-container networking. I'm trying to find how much outgoing data have been transferred but because the month just ended I have no idea how accurate the numbers are.