actually been working with those for some time now, using crawlee + playwrigth for few data scraping projects, mostly large scale ones that needed lot of optimization for cost and speed. i did build my own actor setups that handle pagination, retry logic, proxy rotations and memory tuning, so i kind of get how tricky it can be when site like etsy or linkedin break your runs or show no data. I tried to solve it before using smarter request queue and lighter memory settings, it helped much. also i spent time learning how compute units really work, and from my experience if you manage concurrency and cache properly, the cost goes down a lot. maybe someone here already know same thing, but i found timing and retry delay also matter more than people think. for youtube or instagram part, i did similar scraping for trend or keyword search, tracking posts and engagement too. still improving it, but it runs decent now.