I’ve never really tried to performance-benchmark the OpenAL code.
First, remember that the free version of OpenAL is not hardware accelerated. All that sound-decoding is being done in software. I wouldn’t play more than 4 or 5 sounds at a time.
Second - it supports streaming sounds from disk. If the sound file is larger than a certain threshold, it will play it straight from disk rather than load it into RAM. I believe there is a way to tweak the threshold there. But streaming multiple files from disk concurrently could be an issue.
Third, that python code to update sound positions is… well, it’s in python. How many sounds do you have in the 3D manager at a time?