"There are a number of outbound discards on the two 1gbps port channel uplinks.
Could this be the cause of packet loss and latency?"
As a output discard is dropped/lost frame/packet, yup, it is one (actually the most common) cause of packet loss.
Regarding latency, as output discards are generally due to buffer/queue overflow, such buffers/queues add to latency, but often the impact to the data transfer protocol (e.g. TCP), as it slows transmission rate and/or "recovers" can cause much latency.
"What can I do to resolve this?"
That a depends answers. Another corresponding question, should be, do you need to resolve it?
Generally, solutions resolve around: increasing buffer/queue resources, increasing available bandwidth, "better" traffic flow/rate management and/or treating different "kinds" of traffic differently (i.e. some traffic are very intolerant of drops/latency, while others are the converse - so you might "push" drops and/or latency from the former to the latter).
Lastly, almost any kind of traffic will tolerate some level of drops and/or latency. Impact, again, varies upon type/kind of traffic.
Regarding do you need to resolve, Giuseppe notes, one packet loss per million is not an issue with any kind of traffic. Giuseppe also notes one per one hundred is an issue (BTW, that's an old rule of thumb).
However, IMO, a percentage loss, alone, might tell you little. It's much the same problem as basing bandwidth needs on a utilization percentage. What's often overlooked is the time window the percentage is based against.
For example, say you have 25% loss for 8 seconds. But, if you had no other loss for 4 minutes and 52 seconds (common 5 minute SNMP poll period), you "only" have a loss 0.67%. Is this a problem? Again, depends on the "kind" of traffic.
