Comment by tga_d

Not only would you need the node to expose IPs with a wide enough distribution to allow the right path selection, you'd also need to have enough bandwidth available to look like distinct hosts, and ensure any losses in connectivity aren't correlated enough to draw attention (people monitor metrics.torproject.org pretty diligently, and would notice if there was a chunk of bandwidth coming and going in lockstep). At that point, the difference in cost to just actually running legitimately separate hosts is negligible. All empirical evidence points towards the status quo that has existed for most all of Tor's existence: if you want to identify Tor users, there are cheaper ways to do it than dominating the network (and those ways are expensive enough to be outside most people's threat models).

That said, any bandwidth anyone wants to contribute to mitigate such attacks is always appreciated, even if it's more useful for performance reasons in practice. ;)