So, since I haven’t found a way to filter log, I did the second best thing - I’ve created a workaround.
Every time one of the devices goes to unavailable (for 30s) an automation is triggered that puts that device’s name and current timestamp as a new item on a dedicated to-do list.
And every night at midnight same automation goes through that to-do list, marks all items as completed and deletes them.
In such a way I can always know how many devices were misbehaving on a specific date, and during the day I can check the list of the exact devices.
Oh, and how to clear whole the to-do list I found in this post here.
Maybe it will help someone with the same needs. 