Context Navigation

Changes between Initial Version and Version 1 of Ticket #9111, comment 12

initial	v1
1		Okay, so after a bit more testing, it seems things still doesn't behave perfectly well... We still have situations where the server can't cope with the current load and new threads keep being created faster than already created ones can finish. Sometimes this peak goes back to normal, sometimes it only gets worse until the system resources are exhausted.
	1	Okay, so after a bit more testing, it seems things still don't behave perfectly well... We still have situations where the server can't cope with the current load and new threads keep being created faster than already created ones can finish. Sometimes this peak disappears and load goes back to normal, sometimes it only gets worse until the system resources are exhausted.
2	2
3	3	This is not so surprising, as many times during the lifetime of a request, a thread will do some work with an active connection, then will release it completely and the Connection goes back to the pool. Each time the thread will try to reacquire a Connection in presence of waiters, it'll have to wait a full cycle again i.e. until all the previous waiters have got their turn.