Is it good practice to parallel relay contacts for increased current capacity?

No you should not do this. Sometimes it is explicitly allowed on the data sheet (but not that I can see on this data sheet), and when it is, in my experience you never get as much as double the capacity.

Paralleling physically separate relays is worse again because they're not physically moving together- expect welded contacts etc. if you tried that.

If you can split the load (for example, instead of a 40A heater use two 20A heaters) then you can get an equivalent functionality.

You could think about ballasting the loads (wasting power to roughly equalize the currents) and fusing each contact separately, but I don't think that's a good idea at all.

Note that using the relay at the maximum rated current will lead to a pretty short life (only 100,000 operations for a resistive load), which might be only weeks or months if it's switching continuously. At 3HP (motor load), the life is only 1,000 operations, so at once per minute it won't last a single day.

Edit: With the added information that you're using the relay to switch effectively at a relatively low DC voltage and you're mostly concerned about carrying current.. I can't say categorically this is really a horrible idea with a single relay, but I think I'd get on the horn to the manufacturer and see if it's possible to get any buy-in. It comes down to variability in contact resistance vs. the resistance of the connections (plus whatever, hopefully balanced, resistance you add externally). When one of the contacts inevitably fails first, I think I would prefer the relay to not emit excessive amounts of smoke or flames). I think you're okay at 40A (with AgCdO contacts) given the UL508 rating, but beyond that is in question.

If you really need such a high carrying current, the Omron G7Z appears to explicitly allow paralleling the 40A contacts without derating, for 160A total capacity, but perhaps not with the blessing of safety agencies.

Relay current-handling ability is limited by two factors: the ability of the contacts to pass current when continuously "on", and the ability of the contacts to handle the stresses associated with switching. In general, if two relays are connected in parallel, the continuous current-handling ability of the combination will be almost equal to the sum of the individual relays' abilities, but the switching ability of the combination may be that of the worse relay, and in many cases won't be much better than that of the better one.

If one's application never entails opening or closing relays under loaded conditions, it may be reasonable to use paralleled relays to boost steady-state current-handling ability. In general, however, one should only use parallel contacts to boost "live" current-switching abilities if a relay manufacturer specifically allows it.

Paralleling contacts to pass a higher rating, than the individual contact rating, is a bad practice.

Sometimes paralleling of contacts is done for redundancy. Should one set of contacts degrade current can flow through the other set of contacts. When this is done the normal current load does not exceed the rating of the individual contacts. I've seen this practice recommended by NFPA in Electrical trade magazines where the system is critical and a failure of the relay can lead to another hazard.