A stop loss limits the loss on one trade. Nothing about a stop limits how many losses can arrive in a row, and a drawdown is exactly that. Otto's answer works at the level of the whole account, in two steps.
The first is a reduction ladder. When the account has fallen a set distance from its peak, Otto halves the size of every new trade. The strategies keep running, the signals keep coming, but each one risks half what it did. When the account recovers to its peak, full size returns. The effect is that a drawdown deepens more slowly in its second half than in its first, and the trades that dig the account out are taken at a size the reduced account can bear.
The second is a hard halt. Past a deeper level, Otto stops opening new trades altogether. Positions already open run to their stops or targets; nothing new is added. The halt is the line beyond which Otto has decided that the right response to a losing period is to stand aside rather than to keep trading through it, and it is set below where the reduction ladder has already acted.
Both controls cost something. A halved size after a drawdown means the recovery is slower than the fall, and a halt means some of the trades that would have started the recovery are not taken. That is deliberate. The arithmetic of drawdowns is not symmetrical: a 20 per cent fall needs a 25 per cent rise to recover and a 40 per cent fall needs 67. Controls that keep the first from becoming the second do more for the account's long-run result than any improvement to the entries could, and they are the reason a strategy's owner is still running it three years later.
The thresholds are stated in Otto's documentation. They are chosen from the strategies' own drawdown behaviour, deep enough to be uncommon and shallow enough to matter.