I agree with the main idea: producing new creatives and coming up with very different concepts, frequently, and feed Facebook with enough data to find the winners, is key to improve performance.
Adding new creatives to the same campaigns and let old creatives die their beautiful deaths if they are outperformed, is a fast & efficient way to work, giving you imo 80% of the upside you could get with a cleaner test & analysis. It also you gives you the opportunity to have different creatives perform well together, for different mini-audiences within your big audience.
I find that shipping a ton of creatives, right when the ideas come to you, and empower everyone to do so in a team, creates a happy mess of results & let the truly best ideas bubble up.
Then after 1-2 months, dissect the results, slice & dice every dimension, clean up the mess and restart fresh campaigns with your top 2-3 ideas and a couple of variations. And let it compete with the old, messy campaign, which should slowly decay.
Managing a clean test environment, collecting enough data for a stat sig read, making sure the parameters/variables are the same, analyzing and sharing the results, can slow down the production quite a bit if you don't have a good PM. But if you have really good process in place and it doesn't slow you down, I imagine that's the holy grail.
In conclusion, I'd say that both approaches can work and to pick the one that fits your team style. And also your brand & budget (the bigger, the more structured you'll need to be, approval process etc.)