Polling at different intervals?

You are correct, and it is granular down to the ms.  You can see by looking at the timestamps for just about any data point in the db, I pasted below an example from my lab showing from one poll to the next for a bunch of devices on a 10 minute interval, you can see they are mostly staggered every few seconds, but sometimes clump up.  The scheduler does the polling on a rolling list based on the time a device was added + the polling interval, doing a poll now will insert an extra datapoint into the schedule but does not change your normal scheduled slot.  I think the only way to get your object reassigned in the scheduler would be to unmanage and remanage the node.

Do you show a lot of packet loss on your nodes Antonio?  Orion only sends out a single ping for each up/down poll so if your buffers were being crushed on a regular basis it would definitely show up there.  I honestly would be pretty worried about any network brittle enough that Orion is over running the switches.  In my lab environment running every module with 500 elements we are maxing out at 3 mbps of traffic (including writing the polled data back to the sql server and serving up pages to myself and the other engineers using the lab each day.  If I project that out it would seem that maxing out a primary app server polling 12,000 elements and a stack of SAM monitors and VMAN polling and such would probably move about 70-100 mbps of traffic to monitor thousands of switch ports and servers.

You can see Cisco buffer errors if you look at the network tab on the node details view.  Click on the blue text to see the errors charted out over time.  Check this on the nearest switch to the orion poller and if your problems stem from polling then you should expect to see a nearly continuous stream of misses and if things are really bad then there should also be errors (a miss isn't a dropped packet until it becomes an error)

Hopefully your network devices support QoS so you can prioritize the traffic getting dropped because honestly if you are worried that your monitoring is tipping you over the edge of the port capacity then you will be really jammed up if anyone tried to copy a large file from the file server or print a pdf to a network printer.