Would you run a train network at 100% capacity all the time? No - I don't think you would. There's no slack. If one problem occurs then there's no room to manoeuvre. There'd be a crisis, a meltdown.
In fact - that's one of the problems with London Heathrow airport. It runs at 98% capacity. When bad weather hits, big delays ramp up quickly unless some scheduled flights are cancelled.
So, it stands to reason that as managers we shouldn't try and run our people at 100%. If we do, a crisis will easily happen. We'll then be firefighting instead of growing.
Or - maybe not. The counter view is this: if we are running at less than 100%, we are doing everything with time to spare. Some of that work is less important than the rest. If we say no to things, let's say no to the least important things. If you're saying yes to everything you're not prioritising.
One way to navigate this dilemma is to make a distinction as follows...
Does all the work have to happen within a fixed time frame. Or not?
Predicable fixed tasks with defined time frames - these are best run with some slack capacity to allow for unforeseen circumstances.
Transport and logistics would fit this model. Or fixed deadline civil engineering projects.
Unpredictable tasks with non-defined time frames - these are best queued in order of priority. They can then be worked through at 100% capacity taking a new task from the top of the list as soon as another is completed. If there's a problem, simply pause until the problem is dealt with and then get back to the high priority work.
Software development and creative work might fit better with this model.
Of course, there's no right and wrong here. I'm just highlighting that there are benefits to running with slack and benefits of running with no slack.
Ultimately it boils down to a simple question, "what would happen if we can't complete this by a certain time".
As a startup you're usually running at way above capacity simply because the answer to that question is the amount of cash you have left. At some point however you might need to design slack in your system to avoid a crisis.