video processing – Real-time seeking and playback ffmpeg use case

What is best approach to store the 24×7 RTSP recorded streams and format that can satisfy near real-time seeking and playback preferably without concatenating or trimming them first and then serving the resulting file?

We have attempted using h264 videos that are segmented into 15 minute chunks with a high key frame interval where the user may request for a start and end time that can span either multiple videos or a single video file.

With FFMPEG this is relatively straightforward but we would need to either concatenate or trim files based on the input which takes a tiny bit of time even with copy codec accurately and even on higher-end EC2s thereby leading to a unsatisfactory user experience.

Additonally attempted to use hevc/h265 to also tackle cost efficiency as this is also a factor in the long run but hevc videos are not widely supported on most browsers.