Persistent ALCcontext with source pool versus ephemeral context per playback burst for low‑traffic OpenAL workloads

Goal

Minimize idle resource usage of an OpenAL implementation for a low‑traffic audio workload while ensuring that the latency to start a sound burst remains within acceptable limits.

Constraints and uncertainty

• Persistent context keeps a device handle, mixer thread and source/buffer memory alive, eliminating per‑playback latency but consuming resources even when silent.
• Ephemeral context (open/create/destroy per burst) yields near‑zero idle cost but adds device open + context creation latency and risks device loss between bursts.
• Implementation‑specific factors such as backend‑dependent context creation time, power‑saving policies that may close idle audio sessions, and the lack of a standard pause/resume API affect the trade‑off.

Given these trade‑offs, what configuration best balances idle cost and startup latency for a low‑traffic scenario?

• Is it preferable to keep a persistent ALCcontext with a small source pool and accept the constant idle overhead, or to open and close the context for each burst and absorb the latency?
• How does the chosen backend (e.g., WASAPI vs. PulseAudio vs. OpenSL ES) influence the decision, considering measured context creation latency and idle power consumption?