I sounds like your question is more business related as to which technology you should use. The Intel WebRTC platform is mainly is not a plug and play system that you can easily install and deploy a working solution. With some serious programming efforts you can probably get the Intel library to work as you need it to. Since its not open source, I doubt some of those features you are looking for will be easy to develop.
I expect to do a lot of programming around this solution, but I don't want to start using this if it turns out that any of my final expectations are unrealistic.
1) Is it reasonable to expect that the serverside component of this will be able to mux 26 audio streams together, and grid 26 video streams, and return those back to all 26 participants? Is that within the compute power of a given server, or is it outside the capability of this system?
2) Is it reasonable to expect that the serverside component of this will be able to record the 26 audio&video streams into individual files before muxing&gridding them? Is that within the IO power of a given server, or is it outside the capability of this system?
Thanks for considering Intel CS for WebRTC as your RTC solution candidate. According to your description on the requirement, I think generally Intel CS for WebRTC can meet that target. Intel CS for WebRTC server components are highly distributed with WebRTC accessing, video processing, audio processing, recording, etc. Absolutely it can mux 26 audio/video streams into one and record each individual stream. If you need more concurrent rooms, just adding computing resources as building blocks.