#### Optimizations

A naive implementation would simply regenerate all resources, of all subscribed types, for each client, on any configuration change. However, this scales poorly. As a result, we have many levels of optimizations to avoid doing this work.

$(foreach bin,$(LINUX_AGENT_BINARIES),$(eval $(call build-linux,$(bin),$(AGENT_TAGS))))

# Create helper targets for each binary, like "pilot-discovery"
# As an optimization, these still build everything
$(foreach bin,$(BINARIES),$(shell basename $(bin))): build
ifneq ($(TARGET_OUT_LINUX),$(LOCAL_OUT))

It determines this based on the Workload xDS configuration it receives.
When a new identity is discovered on the node, it will be enqueued for fetching at a low priority, as an optimization.
However, if a request needs a certain identity that we have not fetched yet, it will be immediately requested.

    LD_EXTRAFLAGS="${LD_EXTRAFLAGS} -X ${line}"
done < "${BUILDINFO}"

OPTIMIZATION_FLAGS=(-trimpath)
if [ "${DEBUG}" == "1" ]; then
    OPTIMIZATION_FLAGS=()
fi

time GOOS=${BUILD_GOOS} GOARCH=${BUILD_GOARCH} ${GOBINARY} build \
        ${V} "${GOBUILDFLAGS_ARRAY[@]}" ${GCFLAGS:+-gcflags "${GCFLAGS}"} \
        -o "${OUT}" \
        "${OPTIMIZATION_FLAGS[@]}" \
        -pkgdir="${GOPKG}/${BUILD_GOOS}_${BUILD_GOARCH}" \