This is a potentially very risky approach. I was given the task of having my org stand up a large, complicated system. My engineering team built a system that was failing 2% here, 10% there, 5% somewhere, etc. They built recovery and retry mechanisms for every failure scenario. We ended up with a combined failure rate of 40% that was masking the failures but basically doubled our execution time for the service. I locked down all new development and directed all of our effort toward driving accumulated failures below 5%. They completely disagreed with me (because they'd gotten used to the life of being firefighters). Once we got to below 5% their opinion changed completely. They're very proud of how they were able to drive down the error rate.
Watch out for accumulated error (or even worse--compounded error)!
On the other hand, you can expect that any network based service will fail and need to retry. It's especially important to make software that relies on cloud-based services resilient in this regard.