Thank you for your reply Technonotice.
Well, I have been doing my own research yesterday and today. And so far I have absolutely no idea why is this happening. Box is Redhat 7.3 last updated (heh, guess there will be no future updates
from redhat) stock kernel (2.4.20-28.7) IPs assigned to virtual interfaces eth0, eth0:1, eth0:2, eth0:3, eth0:4, apache is 2.0.48.
I have checked things again with a nice utility iptraf and it confirmed and showed the same difference in bandwidth utilization. When apache extended status shows there have been about 50 megabytes of transfers made, detailed stats of iptraf for the same period of time show only about 5 or less megabytes.
Tried to put squid to control bandwidth for this IP. Configured it (assigned real IP:80 to squid and redirected httpd_accel_host 127.0.0.1:80 then reconfigured apache) and.... I can see absolutely NO difference. Though, if I block everyone (using iptables) from accessing this IP and permit access only to myself I can see that squid is doing it's job well and bandwidth is really limited exactly to the rate I put. But, as soon as I delete iptables restriction rules everything starts again, i.e. download starts to break and apache shows crazy bandwidth when at the same time iptraf shows ~10 times lower traffic usage.
I have just scanned system for viruses and trojans with f-prot and it found nothing, will run chkrootkit additionally, but don't think it will find anything either.
Please, if anyone have any idea why do such things happen, please reply. I am really desperate, don't know what else to do...