I had success with two methods and I am mentioning both of them down below. However, both the methods were tested for 10 days on 3 accounts, all of them run on my home network without proxies (Wifi). So I am assuming that residential proxies would work provided that they are truly residential.
Also, 3 accounts is a very small sample size but nevertheless a good pointer
Method 1: This was tested on one account. No scrappers used.
Settings - Like 1-3 posts per operation. Interval between each operation - 1min. And then Random pauses between x number of intervals. Target sources were hashtags.
Logic: When we open a hashtag and like posts, I assumed that most of us open multiple hashtags and like a few in each of those. So to replicate human behaviour, i did this. I started with 150 a day and increased by 50 odd each day and its at 700 now without blocks.
Method 2 - Like 2-4 posts per operation. Interval between each - 1min. Then random pauses between x number of operations so that the total number of likes for each session is between 25-40 (25 seemed to be a safe number to like continuously without a soft block. Even while doing manually, when i liked posts fast, i counted each time and i got a soft block for a few mins once i liked about 25-30 posts. And this limit seemed to increase gradually. So better to start with about 20 and gradually increase to about 40). Scrapper used. Tested on 2 accounts
This was to replicate how many people manually open links from engagement groups and like them. The actions more or less take the same amount of time or atleast this is the closest that could manage.
Started again with 150 a day, increased by 50 odd a day. 10 days without blocks.
For both the accounts, I dont think its possible to do more than 700 a day due to the speed settings that we use. For some, it might be a limitation. but if we are able to scale this up without blocks, I am sure 700 a day would be like back to the good old days.
Let me know if any of these settings work for you.
PS: I would recommend the second method