First "id est" is used to explain or clarify, it is not a list of literal examples.
I'll again state that I am referring to "tech role" workers commonly represented by an "SDE"; not vendors, contractors, hourly, warehouse associates, etc. I am also referring to this population across all of Amazon; Consumer ("Retail"), AWS, and "Digital" aka "Kindle". I make no specific claims to particular internal organizations which may or may not represent the average.
I don't think that anyone is going to repeat or cite a companies internal data. Maybe we can use public data and some simple deductions to arrive at the most plausible explanation.
Your linked "geekwire.com" article is actually an illustration of the SEC 10-K filings. Under that documents definition of "Employees" it specifically terms this to be "full-time and part-time employees." Contractors, vendors, and temp workers are _not_ included in those numbers: "Additionally, we utilize independent contractors and temporary personnel to supplement our workforce." Unfortunately, that does coalesce "tech roles" with distribution ("warehouse") facility workers etc.
Additionally by reading these 10-Ks you'll note a section on "Stock-based compensation." As many have noted RSUs are a significant portion of Amazon compensation. From 2008 annually this is $275M, $341M, $424M, $557M, $833M, $1,100M, $1,500M, & $2,100M through 2016 Q1. Regardless of employee count the cost of issuing RSUs increased by 25-40% per year, with an increase in rate during recent years.
As you, and others, have noted the Amazon compensation package weights the initial RSU vesting schedule to years 3 & 4 of employment. I believe the typical structure is that only 5% of grants vest after the first 12 months, and an additional 15% at 24 months.
Let us use the public data which shows a distinct increase in employee count and compensation costs starting around 2011. Suppose again the company might have had 5,000 employees in tech roles at that time. Using my posited net growth of ~50% we could expect ~25,000 employees after 4 years, i.e. in 2015. Which happens to align pretty well to your citation of 24,000 in WA as of mid 2015. Don't forget that Amazon is a global company, so while a majority of tech roles may be WA based, it is certainly not all. Feel free to decrease the rate to 40% net, it merely changes the target dates by ~1 year.
We can also compare the public data on employee count and stock based compensation. Using the hypothetical ~25,000 tech role employees for ~2015 and $2,100M of stock we arrive at an average of ~$84,000 vested RSUs per tech role employee. Of course RSUs arent a uniform distribution, but that seems to map to self reported salary data. Adjusting for the employee count at 2015 Q2 and the stock based comp cost at 2016 Q1 further improves the fit.
From your other comment you appear to posit that employee attrition is approximately exponential decay each year. I believe you have also stated that the average tenure at exit time is 1 year. In that case employees should have only vested ~5% of their total RSU grants at time of exit, forfeiting 95% on average. How do we reconcile that with the public data showing a ~40% annual increase in the cost of stock based compensation?
Further your exponential attrition should lead to asymptotic employee counts unless offset. I do not believe anyone is claiming a net decrease in Amazon tech role employees, most cite a rapid increase in active employees. What hiring rate do you posit to support both your claimed 50% attrition rates and the significant net increased employee counts?
PS: AMZN 10-K filings for your perusal http://phx.corporate-ir.net/phoenix.zhtml?c=97664&p=irol-sec...