Whether you're poor or in a lower caste you still get some form of meritocracy within your group to keep you content and motivated. But crossing the boundary to the next level, while not impossible, is designed to be highly unlikely because people in all layers like the security those boundaries provide them.
As people in Silicon Valley have more than 500 USD assets (citation needed), I don't see how this paper applies to Silicon Valley at all. There can be another such threshold, but the paper doesn't provide any experimental support for one.
Whether it's $500 in Bangladesh where Google tells me the "typical" salary is ~$150-$300, or $500k in SV where the typical pay might be $150k-$300k the principle is still very much the same.
Everyone claims meritocracy but in the end it's proven it's not individual merit we're actually looking for. We're looking to maintain barriers around our layer of society and allow some form of real meritocracy only within a layer.
If you don't have access to schooling and someone else does then that's something that will make you always start with a handicap when competing with them. The boundary, whether education or medical services, is there so one layer feels safe from the inferior one. People want it there even if this means they start off with a handicap compared to the superior layer. Most people are more afraid of losing what they have than they are keen on winning something they don't already have.
"Evidence on Poverty Traps from Rural Bangladesh" - are you suggesting the situation in rural Bangladesh is applicable to Silicon Valley?