|
|
|
|
|
|
by jacques_chester
4850 days ago
|
|
|
The documentation used to say that requests were routed to the next idle dyno, but the behaviour was silently changed to be random distribution, which is substantially worse for the customer. There seemed to be previous cases where Heroku were aware of the discrepancy but didn't act to fix it. That's generally considered uncool. |
|
|