edit: This is the general approach by US agencies, the IRS website barfed on my info and I had to call a local office to get a person to help me (the nation wide number was 100% automated and likewise barfed on my info).
> Laid-off workers are confused and confounded by the department’s faltering claims system, erroneous denials and stubborn silence on key policies and questions.
> The frustrations lead people to call the department again and again – some say they dial more than 700 times a day.
> The employment department is taking steps to address the call volume. It has added hundreds of staff to process claims in recent weeks – it has 520 now and has leased a facility in Wilsonville to expand claims staff to 800.
> With jobless claims up nearly seventeenfold, though, the staffing increases aren’t close to keeping up with demand. So it may be weeks – or months – before Oregon works through its backlog in claims questions.
Conventional wisdom these days is that if you want to collect unemployment, you need to make filing a claim your full-time job. You start calling at 7am when they open, and keep re-dialing until you get through -- hopefully before they close at 4pm, and you have to start again the next day.
Also, don't call before Wednesday, if you don't absolutely need to. Monday/Tuesday are unofficially reserved for people who really need the money.
Of course, even if you get through to a person and get the right bit flipped in their database, there's no guarantees. I also hear lots of stories of people whose claims were approved 4 or 6 weeks ago and still haven't gotten a dime.
After a few months, the issue disappeared as mysteriously as it had appeared. And we'd had that set of static IPs for about 5 years by that point.
While I agree that it's unlikely that they would serve NY customers from Romania, you will definitely see weird results from Geolocation services when going through some VPN providers.
That's a bizarre capability for a system that is as incompetent as has been seen.
I wonder if CDNs can be used as "standard candles" for in-page scripts to measure latencies to multiple known network locations and infer a position from there across multiple sessions with different exit points. How many exit-point/latency pairs would I need to figure out the actual origin of a request?