Newbie asking: Are there any hidden limits on the 'unlimited' plan?

I've been evaluating PlayHT for generating system alert audio snippets and on-call runbook narration. The "unlimited" plan looks attractive for our volume, but my SRE spidey-sense tingles at that word. In my experience, "unlimited" in SaaS usually comes with some form of fair-use policy or hidden throughput limits.

I'm looking for concrete experiences, especially from anyone using it for similar operational audio. Before I push to get this approved for the team, I need to know:

* Is there a cap on concurrent generation jobs or a max audio length per file that isn't front-and-center?
* Are there any undocumented restrictions on batch processing or API rate limits that could throttle us during an incident when we might need to generate multiple alerts quickly?
* Does "unlimited" apply to all voice models, including the premium ones, or are there generation credits at play?

I'd rather find out now than during a Sev-1. Any war stories or config snippets showing actual API limits you've hit would be incredibly helpful.

For context, here's a simplified version of the script we'd run, which could be fired off multiple times in short order:

```python
import requests

def generate_alert_audio(alert_text, voice="michael"):
payload = {
"text": alert_text,
"voice": voice,
"quality": "high"
}
# This would be called per alert channel
response = requests.post(f"{API_BASE}/generate", json=payload)
return response.json()
```