Assess whether a FIFO vs LIFO priority queue at broker would harm/help proxy diversity and enumeration

Some UCSC researchers doing measurements of the snowflake proxy poll brought to my attention that the broker's proxy heap implementation functions as a LIFO priority queue (where proxies with fewer clients are prioritized).

My intuition is that this will cause proxies that poll more frequently (such as standalone proxies) to be matched disproportionately more frequently than proxies that have longer poll intervals. I'm also not familiar with the implementation of the priority queue aspect of the container/heap package we use.

This is a good candidate for a shadow experiment: to see how modifying the priority queue implementation affects proxy enumeration attacks and proxy diversity.