You should not assume that you can reasonably control your assets in moments of volatility. You may not be able to control them at all.
Market access is a major component of liquidity. If your broker already holds a phone in either hand, he won't be able to pick up your line when the light next to your name flashes.
Imagine that you have all of your liquid net worth in GLD and a huge cholera outbreak hits lower Manhattan. Bad shellfish. GLD rallies and you want to close your position and take profits. The problem is, the largest firms that make markets electronically on GLD have sent their employees home and shut down for the week. Their absence from the market makes it difficult for you to get a fair price for your large position.
Imagine that you use Verizon for your mobile phone and internet service. You think their service is great, so you buy shares of VZ. One day, Verizon announces that due to a malicious hacker, their network is experiencing cascading nationwide hardware failures that will not be repaired for at least a week. VZ stock plunges, but you can't modify your position because you can't get on the internet or make calls.
A less fanciful example would be owning stock when trading has been halted by the exchange. Perhaps there exists a buyer during the halt who would be willing to trade at a price that is favorable to you, but that transaction cannot happen.
Tail risk, in some cases, is a scenario where stocks get cheap. In other cases, stocks can't trade at all. (And in the worst of situations, tail risk is a scenario where guns get expensive.)
I'd consider all of those example cases where it's not necessarily reasonable to control your assets. I will say the Verizon case is a bit fanciful because it should be pretty easy to hop on another network quickly for most people to bypass that small hurdle...
When we're talking about an enterprise scale electronic trading system that should already have tested these sort of load scenarios, especially in a world where AWS/Azure/etc. exist and elastic computing is the target roll out for applications that probably don't even need scaling, that's not a reasonable case. That's a case where something that should already have been planned for and tested thoroughly slipped by. It's especially damaging when it happens repeatedly over a short time span. There's nothing unpredictable here. Management and IT at Robinhood surely read the news and knew there would be massive load today.
Surely, someone designing systems and software in this domain understand that the market could be highly volatile and their system needs to be ready for those cases where it's flooded by customer requests or clearly conveyed to customers that it's not ready for such cases.
Robinhood operate on razor-thin margins and if you just look at where three-month implied vols were trading in early Jan, we are in a 7-sigma event. To be fair to Robinhood, it is probably hard to anticipate and prepare for such an event when you're not charging commissions. They are not the only brokerage that has had problems in the past seven days. And they lose money when their customers can't place orders.
"You should assume you can reasonably control your assets at any given time." Sometimes you can't reasonably control your assets, because you can't control them at all. So you shouldn't assume that you can reasonably control them in those times.
I hope it is clear that I am not trying to be argumentative; I actually think it's a complete let-down that Robinhood has left so many people frustrated as their wealth whip-saws around in the market. I'm just saying that tail scenarios don't usually fit into our predictive model of what's "reasonable," even when the solutions seem crystal clear in retrospect.
On a more general point, I never understood why market returns are assumed to be gaussian - these "one in a billion" events seem to happen far too often.
Let people sell, that's on them.