Were social interactions unimportant this would be practical, but the value of upvotes isn't only drawn by content quality. Society is based on relationships, and upvotes reflect social values other than content quality.
I also have friends here on Steemit, and of course I am interacting with them and upvoting them. However, there isn't anybody I am upvoting ten times a day, there isn't even anybody I am upvoting five times a day. That's just not necessary to value their content (actually none of them is writing so many articles per day).
Nevertheless there are users who are upvoting own content (or the content of their sockpuppet accounts or their circle-voting friends) ten times per day. Only they would be impacted by a reasonable implementation of diminishing returns.
Folks I follow, @everittdmickey comes to mind, that do publish multiple posts per day, might receive multiple votes from me in one day. I often don't upvote all their posts (I do not autovote, and read most posts I upvote), but folks using autovoters might. @everittdmickey doesn't follow me, so this isn't an example of circle-jerking, just that I find good science and points in enough of his posts that I do upvote more than one a day sometimes.
Then maybe we just evaluate things somewhat different (that's no problem of course). I see no problem to upvote even best friends only two or three times per day, and spread ones remaining votes for example to some newbies who really need attention and motivation ... even with diminishing returns you still could upvote everybody as often as you like, it's just that every upvote would be significantly weaker than the previous one.