QCA does not care about FOSS. I'd personally rather purchase hardware that might not be as performant (consider how much of the peak performance you will need?) but in the end does not depend as much on binary blobs as ath10k/ath11k does. That pretty much leaves MediaTek (or outdated 802.11n ath9k solutions). Ath9k was an anomaly where the company grudgingly adapted a FOSS driver that was (ironically) better than their in-house proprietary code. Ath10k and ath11k are QCA trying to profit off the goodwill that ath9k gained them (none thanks to themselves, ironically). While MediaTek is far from perfect either, at least they seem more willing to share details on their hardware and there is a major OpenWrt developer working on the code of the wireless driver. Not a bunch of QCA minions that only do their master's bidding.
People have been arguing that MediaTek 802.11ax solutions aren't as powerful as QCA, but that point is geting moot now that MediaTek ARM 802.11ax solutions are coming to market and getting supported by OpenWrt. Besides that, people lean towards 'let's get the most powerful device I can get' then forget their clients are often mobile devices (smartphones, tablets, laptops) that only rarely come equipped with outsized radios. Most clients have two-stream radios and that's it. You won't benefit from a router with four streams in that case. But it's easy to be fooled by numbers.
My setup (a multi-storey house) is a wired ARM64 box for routing and multiple MediaTek 802.11ax access points. I have a Netgear WAX202 and a TP-Link EAP615-Wall. Those offer two streams maximum, but with good signal I am seeing over 600 Mbps throughput with iperf. So I am a happy camper.