If the IP addresses get associated with to many spam reports, unsubscribes etc the gmail etc will start to associate these IP addresses as bad, and will be more likely to just block them no matter what the content.
There are actually 3rd party services that mail companies can use to check how trusted an IP is. So a lot of times, they'll provision an EIP on AWS, run a check against that service, use it if it's clean, and release and try again if it's not.
It does lock you to using EIPs though, which makes it a bit harder to scale up.
You might be interested in a provider-agnostic email API I am building to to avoid EIP/provider lock-in and have the security of multiple dedicated IP providers. Check it out: flutemail.com