1. Wrong data
We see a very large and unreasonable render_latency reported from WebApp.
+----------+----------------+
| os       | render_latency |
+----------+----------------+
| LINUX    |  947.377375928 |
| MAC      |  8892.83722863 |
| WINDOWS  |  9668.91868007 |
| CHROMEOS |  1756.90810538 |
| ANDROID  |  32.3396733496 |
+----------+----------------+
During last 10 days, we have in total 174M records, and 1.3M of them are larger than 1s, 0.96M are larger than 10s. More specific, there are 3596 records equal to 90380560, which does not like to be a reasonable value.

2. InaccurateSpecifically latency values we get for FPS below 10 become more significant (they are counted multiple times) compared to the same values when FPS is above 10 (many samples are essentially ignored). 
WebApp calculates the average in last one minute, instead of last * samples, which makes some samples become more significant than others. Specifically latency values we get for FPS below 10 become more significant (they are counted multiple times) compared to the same values when FPS is above 10 (many samples are essentially ignored).
 

1. Wrong data
We see a very large and unreasonable render_latency reported from WebApp.
+----------+----------------+
| os       | render_latency |
+----------+----------------+
| LINUX    |  947.377375928 |
| MAC      |  8892.83722863 |
| WINDOWS  |  9668.91868007 |
| CHROMEOS |  1756.90810538 |
| ANDROID  |  32.3396733496 |
+----------+----------------+
During last 10 days, we have in total 174M records, and 1.3M of them are larger than 1s, 0.96M are larger than 10s. More specific, there are 3596 records equal to 90380560, which does not like to be a reasonable value.

2. Inaccurate
WebApp calculates the average in last one minute, instead of last * samples, which makes some samples become more significant than others. Specifically latency values we get for FPS below 10 become more significant (they are counted multiple times) compared to the same values when FPS is above 10 (many samples are essentially ignored).