Disclosure: I work on Google Cloud.
For what it's worth, I often recommend customers look to our partners (e.g., Elastic with their managed offering). The people who build the product will know it best. Moreover, they would be way better suited to running say a multi-cloud variant, or staying with you as you move between cloud providers.
There are still barriers though for third-parties on a cloud platform, some of which we as providers can slowly chip away at:
- Integration/Feel: You want to be in the console and feel like a natural part of the platform. This is double-edged though, because holding all partners to such a high bar (say a Google internal API review... ha!) is unrealistic.
- Support: Who do customers call? Their overall cloud provider or their specific service provider? If the customer is having trouble with their Elastic cluster, is it because we're having a GCE outage or did the customer write a query-of-death? For the latter, our front-line support won't be much help. [Edit: And you want to be able to still have a very fast "Hey! My stuff is down!" path. I think technological things could make automatic handoff here much better, but I believe it'll be "not great" for a while.]
- Economics: most service providers are going to be smaller than our largest customers. Those customers have usually signed up for multi-year commitments with high volumes resulting in a discount. If your service pricing model is "on-demand GCE pricing plus our service fee", you suddenly look over 2x as expensive to their TCO folks (just from missing Committed Use, or RIs with AWS). Unbundling the resources from the management fee would make this easier, but I've not seen many service providers attempt it.
We owe you much better docs (start here [1]), but we're trying to get there on at least some of these. All of the "human" factors though are still going to be a challenge, no matter how much software gets written.
[1] https://cloud.google.com/service-infrastructure/docs/tenancy...