FAL.ai Concurrent Request Bug: Stuck IN_PROGRESS Slots With No Fix — Developers Are Switching

On March 24, 2026, a developer opened GitHub issue #939 on the fal-ai repository with a critical reliability report: after revoking their API key on fal.ai, 6 requests remained stuck in IN_PROGRESS state and continued consuming concurrent request slots. The problem? There is no self-service way to clear these stuck slots . As of this writing, the issue remains open with no fix deployed . Why This Matters for Developers Concurrent request limits directly affect throughput and billing. If stuck slots cannot be released, developers face: Quota lockout - you cannot make new requests because ghost slots are eating your limit Billing confusion - stuck requests may continue to accrue charges Security risk - teams that rotate API keys for security reasons are especially vulnerable In production environments, this kind of cascading failure can take down your entire pipeline. What to Look for in a Reliable AI API Before choosing an AI inference provider, ask these questions: Does revoking a key