Everyone I knew who got LR'd was regularly recognized for the hard work they did above & beyond their regular duties.
Go to a small systems house which works effectively on fixed priced contracts with major penalties for late delivery and you'll see a level of game far above what you'll see at Cisco even from the people with only 1-3 years experience.
I worked with many of the highly acclaimed on the engineering side of the house, and while they put in massive hours and were highly acclaimed they didn't have even the most basic skills. I'm talking elementary school skills like reading, writing and independent learning which is why everything was poorly shot from the hip, not just things you would get in a good undergraduate program like "what are requirements and design and how do they differ," functional decomposition and refactoring which few TLs and zero PEs I worked with there understood. Many once had the potential to be excellent engineers but like frogs slowly boiled in pots they weren't aware of what happened.
I can point to both hardware and software programs which shipped over three years late because they had to be redone repeatedly while the causes were easily preventable. Considering some on the software side were 1-3 month programs those are big misses not even counting the massive increase in body count after the programs were declared "complete" at that 1-3 month mark. The ADHD part comes from those years of post-completion tasking either being 23 hour often useless but measurable tasks between one hour daily meetings or 56 hour bug fixes, three of which can be disposed of in a 168 hour or seven day period (I've been there when directors threatened to fire people for saying "week" in meetings.)
Can anyone name another company where nightly builds are commonly broken for weeks to months at a time and branches sometimes abandoned because they are declared "unfixable?" This happens because the dashboards reward check in rate but don't track the corresponding build breakage so there is a strong disincentive to take the time to get it right in one shot. This cascades to the point where it's blindly accepted that this is how development is best done, which gives people the wrong perspective on what is actually effective.
On the hardware side a hypothetical 1THz processor with no cache and a narrow 1MHz memory bus for all instruction and data access isn't "fast" because of the 1THz processor - it will be stalled almost all the time. There are those rated as "best and brightest" at Cisco who don't get the fact that more than one component can limit the performance of a system and a number of designs ended up being redone over this. Likewise on the software side high schools in the 1970s taught people how to factor common code at the design phase to use GOSUB in BASIC but the amount of replication and 1,000+ line functions in the major code bases at Cisco is ludicrous. How many bugs do you have to fix when n bugs are replicated (by both independent development and cutting and pasting) across m dimensions (code in functions, functions in files, files in directories, directories in trees, and trees in branches?) This is how you end up measuring bug counts in blocks of 2.6 million.
Now that Cisco can't afford to waste money like this the fingers are being pointed at deadwood but they aren't responsible for any of these things. Who does that leave? Hint: it's not just "the other."