+1 to Oren's comments above on polling within an asset if it's a relatively short wait, or using a sensor if it's a longer wait.
One thing to note is that if you poll within a sensor instead of within the asset, downstream dependencies of the asset may not work as expected (since the downstream asset will believe the upstream asset has already materialized, so thus the downstream asset will begin execution). So you may have to define downstream assets in a separate job and kick off those assets using the sensor.