I can say a bit as I'm now retired.
I ran a portfolio of fixed income systematic strategies, mostly trend following. To give you an idea of size, if the entire portfolio was in 10 year treasuries the peak position would have been around $25bn.
Given you're mostly in trend following you would be gradually adding to positions, and then gradually removing them, but you are constrained by the cost of trading that market as to how fast the trends you catch would be. Generally you'd set a limit on how fast you're prepared to trade depending on the cost of the market. In the most liquid markets it's probably okay to trade every week. In less liquid markets, like credit or EM swaps, you might only be able to turn over your portfolio a few times a year. Getting this right is essentially a statistical exercise of comparing a known cost with an unknown payoff from faster trading.
You want to be able to get out of a position in a relatively short period of time, without impacting price too much. So you might set a rule of thumb like 'I want to be out of my position in five days without being more than 10% of average daily volume'. That implies you can't hold a position of more than 50% of ADV (or perhaps less, given that volumes will probably fall in a crisis). You also might want a limit on what proportion of the open interest you are.
So the larger the portfolio you have you have to reduce diversification by sticking to the most liquid markets. This comes down to setting limits on any portfolio optimisation you have.
Interestingly the total costs you face at that level are fairly close to the costs as a 'point and click' retail trader (trading futures through IB at least). There is a sweet spot for costs for funds of around a billion dollars (depending on the asset class) where you can get institutional commission levels, but not be impacting the market. So the game isn't as different as you might think.
