[pranavv00]

Summary

Fixes a race condition where concurrent async calls could trigger multiple redundant token refresh requests.

When multiple coroutines executed simultaneously (e.g., via asyncio.gather), they all passed the token expiry check and independently called the refresh endpoint, resulting in N duplicate refresh requests.

---

Solution

Introduced an asyncio.Lock to synchronize token refresh:

• Ensures only one coroutine performs the token refresh
• Other coroutines wait for the refresh to complete
• Added a double-check inside the lock to avoid unnecessary refresh if another coroutine already updated the token

---

Example (Before vs After)

Before

await asyncio.gather(*[admin.a_get_realms() for _ in range(10)])

→ 10 token refresh requests ❌

After

→ 1 token refresh request ✅
→ Remaining calls reuse refreshed token

---

Safety & Scope

• ✅ No changes to public APIs
• ✅ No changes to existing behavior (except eliminating redundant refresh calls)
• ✅ Fully async-safe (no blocking operations introduced)
• ✅ Minimal and localized change

---

Notes

This follows a standard concurrency control pattern for async systems to prevent race conditions during shared resource updates.

Happy to adjust if there’s a preferred synchronization approach.