I've been battling nearly the same issue for years. I work with antibodies and antibody conjugates on silica based HPLC columns was well, and have always been disappointed with the number of injections I get before the peaks become too broad to work with. On average I may get 250-300 injections out of a column, and I think ONCE came close to 500. What's strange is, if I run the manufacturer's QC test (usually something like PABA, or some other small molecule to determine the plate count), on a brand new column and a column that is "dead", I get the same result. The peak broadening only seems to apply to the antibodies themselves. I have found a few things that seem to help a little:
1. ALWAYS filter the mobile phase. We use a totally aqueous mobile phase, with sodium azide added to retard microbial growth. One thing to be aware of with sodium azide however, is that it is not stable at acidic pH. Silica however, doesn't like high pH, so you're stuck between a rock and a hard place. Also, if you adjust the pH of an azide containing mobile phase using an acid, add the azide last, as when adding the acid you create local "pockets" of extremely low pH, and some of your azide decomposes.
2. If you're working with antibodies/conjugates from multiple hosts, watch out for situations where, for example you inject an antibody from a mouse host, then later inject an antibody directed against mouse immunoglobulins. I don't care how inert the manufacturer claims the stationary phase is. Some of your sample is always going to remain on the column, and your next sample may then interact with the previous one, and so on.
3. Some manufacturers recommend that before using a column, you inject a few large slugs of some "inert" protein such as BSA to "block" any potential sites on the column that might interact through some mechanism other than size exclusion. I don't know how much good it does, but if I get a new column and do several injections of BSA, I do see the peak area increase slightly with each injection until it finally plateaus. There are some mobile phase additives which have been shown to reduce secondary interactions.
As far as cleaning goes, I too am guilty of waiting until I notice degradation in performance to clean (low pH/high salt, organics, guanidine, urea, etc). It hasn't worked once. Perhaps if I cleaned more often it might show some benefit...
And finally, I do use guard columns (cartridges) in addition to a pre-column filter, even though I filter all samples. I can see on the cartridge when I change it that it usually catches something. Either particles shed from the rotor seal, or in the case of conjugates, the color of whatever conjugate was injected most often.
I agree that it's usually best to just bite the bullet and buy a new column, but I wonder why SEC HPLC columns seem to be more expensive than, say, C-18. If someone could shed some light on that, I'd appreciate it.