Ask HN: Mitigating risk of using AWS/GCP/AZ/AC specific platform services

To the engineering teams on HN: How do you mitigate risk of using AWS/GCP/AZ/AC specific platform services in your own productively used applications/services?

There are valid reasons, which will require moving your productive app across other cloud provider. Hopefully, without business impact, in trivial timeframe and minimal costs.

Use of computes across the vendors is largely simple. I want to be prepared for situation when I use specific services which are not equally spread out in other vendors.

Multi-cloud abstractions start becoming costly and can be too restrictive.

Keen on hearing HN experiences on this.