Distance, so a separate building far enough away from anything you want to keep that it won't carry over. Cabling underground but not in a duct. Space the cells a bit and ensure good airflow between them. A single runaway cell is how it usually starts and that in turn usually first shows up as increased internal resistance. Avoid mixing capacities, brands, different internal resistance values, cells that self-discharge, cells that have any sign of physical abuse or damage. (The guy in the video totally misses the internal resistance angle, as well as the self discharge which takes a long time to test and he already spent two months or more on his power bank.) Weld, do not solder your connections. Assemble cell groups first and then do QA on the cell groups as if they are larger capacity cells, and monitor them with a FLIR for any sign of cells misbehaving, especially under larger cell group charge/discharge currents. If a cell in the center is more than a few degrees warmer than those on the outside that's a serious concern.
Other prevention measures: strict inbound QA on the cells.
That's a lot of work, to the point that if you value your time you are better off buying a factory made pack of a reputable because they will almost certainly do a better job than you will on the safety aspect.
Getting Lithium-Ion packs right (especially larger ones) requires more up-front funds in terms of gear (especially testing gear for large volumes of cells is quite pricey) as well. The only reason I would do another (big) pack is if the form factor or capacity I want is not available at all.
The brief flash of that 'rechargeable powerbank' in the video has so much wrong with it that it isn't even funny and that's before looking inside the pack. All of those crossing wires, brrr. And those modules in the linked video don't look much better. Oh, and he's got one cell group in there with fewer cells (13:26) so that pack will unbalance immediately during charge or discharge. Effectively the whole pack has as much capacity as that one smaller cell group times the number of cell groups. The lack of integrated balancing wires is another puzzle for me, crossing balancing wires is a major headache when building larger packs, you want the very best grade of wiring for that with vibration resistant insulation and some kind of wire guide to ensure the wires can't move or cross. The whole BMS setup looks like an afterthought, rather than that it was designed in.
Oh, no interlock on the two separate breakers for the inverter (configured to work in island mode) and the house power. Wait until someone engages both. You need a transfer switch there, not two separate breakers.
His remark that it all looks 'super dodgy' and that 'I do not recommend anybody ever builds a pack like this' is an interesting one: if you are aware of all that, why do such a crap job in the first place? If you are going to go through that much work you might as well do a proper job.
Meanwhile, this is really just an ad for JLPCB. Running around the house and the workshop at everything working is a bit cringe, it is as if that's just padding to extend the video runtime.