Live stream dies only page reload, only more Than one viewer

I am experiencing some odd behavior livestreaming and hoping the community may be able to spot the issue or know the source, I am a little stumped.
Ubuntu, Docker, Caprover as a PaaS PeerTube latest, livechat latest is and is not installed during tests with no difference, OVH S3 object storage for videos and playlists, no CDN (at least till I find one cheaper than serving from OVH heh)
I start up a livestream, 720p encoding settings 2 keyframe interval around 3500 kbps nothing too intense. Everything is all fine chat works, stream federates, cross instance chat federates until I reload either local or federated video page, then the stream spins and never recovers across everywhere.

I get a warn in PT logs:

warn[7/5/2023, 10:17:35 AM] Client log: HLS.js error: mediaError - fatal: false - bufferStalledError
  "tags": [
  "username": "root",
  "userAgent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/ Safari/537.36",
  "url": ""

No console errors regarding the stream are in browser (chrome, some unrelated form field warnings on some instances but seem unrelated.)

The Server resources barely nudge while I am streaming so I don’t think hardware resources are an issue. I’m curious is anyone has encountered something like this and perhaps has some clues to a resolution. I’ll be happy to drum up some deeper logs I am just not sure where to start.

Update, I did add a CDN for stored videos only though, they do not support streaming and setting the base_url for the streaming playlists in the config just make things worse. Also I took a shot in the dark, and played matching, my object storage is 3gb max, so I set the following in my NGINX:

  location = /api/v1/videos/upload-resumable {
    client_max_body_size    3G;
    proxy_request_buffering off;

    try_files /dev/null @api;

I now have zero issues streaming. /shrug.