AFAIK there's no easy way to always, securely detect premium numbers. The fraudster can setup a forward from a normal number to a premium one anyway. So checking number prefix is useless. One could listen for the "After the beep, this call will be billed at X c/min" recording that premium numbers in honest countries have. The fraudsters manage to find less honest premium number providers that skip these.
This is a well enough known issue that Twilio has a page about it: https://www.twilio.com/learn/voice-and-video/toll-fraud
If Twilio wanted to stand up for their customers they would refuse to pay for any premium number that's not in an ITU registered premium range. It will cause a few lawsuits with telecom carriers, but then probably ends this problem once and for all.
Twilio claims this isn't possible, but most mobile operators offer an option to disable calling premium numbers. So either the providers are very customer friendly and eat the cost (unlikely, it would open them up for huge amounts of fraud) or the mobile operators figured out how to block it or push the cost back to the upstream provider.
Would the forwarding party not have to pay the premium rate part in this case?
Doesn't this mean that your call from Twillo to the normal number would be charged at your usual rate, and the fraudster would pick up the cost for the forwarding?