How to use MemoryStorage (mainly for RequestQueue) on the platform?

My actor runs typically use Cheerio, take <20m and have around 1k requests.
For this scenario, costs for RequestQueue writes/reads are often higher than compute units. I wanted to experiment with using in-memory storage to optimize costs (i think I understand the associated risks and I'm ok with them).
I've tried setting

storage: new MemoryStorage()

storage: new MemoryStorage()

in Actor.main second argument as noted in docs & TS definitions, but actor runs on platform still seems to use "platform RQ", not "in-memory one". Any pointers?

https://console.apify.com/actors/64sLcqgxq4IB5hZrI/runs/QIkqpBDa846Ftn5xK#storage