I was trying to understand the efficiency and when to use EIFS, maybe one of the experts can help me further.
EIFS is used when a client demodulates the frame and checks the FSC of the frame. If the frame is corrupt it will backoff for EIFS (about 160 microseconds for OFDM). The process is done for all clients regardless of the RA address in the frame.
Now clients don't use the same data rate and modulation. In 11ac Client1 can use MSC9 and Client2 MSC5. C2 will always mark C1 frames as corrupted and backoff for EIFS. In addition for high dense implementations a channel has to be reused (especially in the 2.4G). This will lead to a significant portion of the clients will not be able to correctly demodulate the frames of other "far away" clients and hence backoff for EIFS. This is totally inefficient (the CWAP Book describes this as Near/Far problem).
I noticed that almost all chipsets use RTS/CTS (also for 20MHz channels) before sending 11ac data. As these packets will be sent with basic data rates, this will solve most issues. But as soon as an administrator changes the minimum basic rates for an SSID for instance from 1 to 12 in the 2.4G band. Far away clients on another AP and in the same channel will not be able to demodulate the RTS/CTS frame sent in 12Mbps (or maybe even at higher rates).
A better method would be, wait SIFS and look if another frame has been sent. Then use CCA to read the PLCP header which will contain the frame length, if this length=ACK (ACK Length is known). Then we know the frame was not corrupted and EIFS backoff will not be used. Make sense?
Does the client still really use this inefficient method? Is there an way to check on a client (of whatever OS) how many times it encountered corrupted packets in the channel?
To answer your last question, I know that some chipsets keep count of (some of) these events in their internal registers. Others can be deduced from these counts.
I have only seen this data from specially compiled firmware, and the results had to be read using an alternative interface - i.e. not over wireless. Obtaining the firmware additions from the chipset manufacturer usually requires NDA's and big bucks - say $30k or more..
Being able to get these statistics can give you a finer view of what is actually happening inside your radio, more than just counting ACK's. Some diagnostic hardware, like Anritsu's WLAN Test Sets, can make use of these parameters in what they call "direct" mode.
Personally I prefer data that can be determined in "Network" mode, where the radio looks like a normal STA of some kind.
I do not work with it, but you may be able to find more information regarding your question by studying Mesh networks. They have to play many more "games" with their timeouts, etc.
Good luck. Please share any discoveries you make in this regard.