I still think 2k blocks is too small and I think the increasing of min_signed_per_window should go along with the increase of blocks_window, so that there would be still the same window of allowed downtime but validators who have bad hardware would get kicked out faster.
Don’t forget that a jail for a validator is quite a dirty spot on a validator’s reputation regardless of whether it was slashed or not, and sometimes 2000 blocks is a waaaay too small timespan to fix their node, and sometimes it may result in a validator trying to fix their node by switching to a failover node, resulting in a double sign if the older node goes back online (and yes, I know such cases), and we do want to avoid such cases, don’t we?
Only increasing the min_signed_per_window param would make the life of a validator way more stressful, and I see a trend of having quite a big signed blocks window and overall the time a validator is allowed to skip blocks without getting jailed, and I do not see a reason to make it different here.
IMO the ideal approach here would be to save the same 16 hours downtime window, so if min_signed_per_window would be 0.8, the blocks_window should be 47500 (can be rounded up to 50k I guess to make it easier to count, that way a validator would be able to skip 10k blocks without getting jailed).