> Are you sure you notice a difference of 0.05 seconds?

Yes, actually. I wouldn't notice a difference of 0.05 seconds in
something that takes 2 seconds to run, but I certainly will in
something that takes 0.5 seconds to run. That's 10%, after all.

There's also the fact that 0.5 seconds is such a short amount of time
that I really can't go and do anything else with my brain while I'm
waiting. So all of my attention remains focused on the task at hand,
namely waiting for Emacs to start. This is another likely explanation
for why I notice such a short time difference.

> Are you sure you notice a difference of 0.05 seconds when the
> variance is around 0.15 seconds anyway?

Yes, in fact, I am.

> I can just use Emacs in a tmux session and a terminal, connecting
> with ssh to the server.

If you are using emacsclient, then I am pretty sure you are using the
Emacs server. The server does not have to be headless -- you can start
it from an otherwise normal Emacs instance.

The Emacs server does not work for my workflow even when I run it on
the same machine that I'm connecting from. Introducing network latency
into the mix would just make things even worse.

> ssh -Y server
> emacsclient -nc
> emacsclient -nw
>
> How is this difficult?

In my experience, my color theme is totally broken when I connect to a
graphical Emacs server with a tty client, although it works fine in
both graphical and tty Emacs instances individually. That is what I
was referring to.

I know of at least one other person who has extensive Emacs experience
who does not know how to work around this problem, so it does not
appear to be trivial.

> no fair comparison when you can't let it run

I am sorry but I do not understand this sentence.

> How do you even measure such tiny amounts of time?

With the time utility supported by most shells.