I've found it works decently so far - albeit it's a little quirky.

At least here on OSX, if I don't SSH into a server for 10 mins+, the first login attempt might throw an error, then by the second or third time my Trezor (or the agent) is finally awake and actually responding properly to authorisations.

Once whatever fell asleep finally wakes back up - it generally responds quickly when I attempt to SSH into a server that uses my Trezor key for the next 5 minutes or so, until the agent / Trezor falls asleep.

My guess is it might just be a bug with the macOS SSH agent not handling Power Nap properly - OSX has this "power nap" feature which takes away CPU resources from any application it believes has been idle for too long - or out of focus etc.

Nevertheless, it works *most* of the time, and the concept itself is really great for high security servers. It's also still more convenient than an encrypted SSH key where you have to constantly enter your SSH key password to unlock it etc.