Most of the issues as I see them are more a matter of consistency of detail and coverage. Depending on your area and purpose, they may be a non-issue for you.
From downloading the NHD in Oct, 2007 (shortly after it was anounced they were 'complete) and using them for a 50 State mapset; spot checks of Oct, 2008 data; and some info from the April, 2009 NHD Stewardship meetings:
In a number of locations, polygons for the ocean and some major lakes are missing. If a few cases the lines for the shore are also missing. A major issue if it occurs in your area.
In some areas, island areas were not 'cut out of' a ocean or lake polygon.
Polygons for some wide sections of rivers are missing.
Large rivers (those shown as polygons) are not named; and can not be as any large (two shoreline) stream/river in that subbasin is also part of that polygon and hence the name would also be applied/shown for the side streams.
As each USGS 7-1/2' topo 'stands on its own' (with possibly many decades between production, source materials, etc.), the downstream (and usually larger) portion of a 'minor' watercourse may not be continued on the next quad.
Names were taking from the GNIS with a very small tolerance for coordinated differences. Thus only about 1/3 of the waterfalls in Colorado with GNIS names have names in the NHD. The coordinate matching was very exacting, but the feature type seems not to have been as the NHD coordinates for Lake Powell (reservoir) matched Teddys Horse Pasture (flat) in the GNIS and THP was applied to the NHD polygon.
Some featues are miscoded as to type - also occurs in the Census data.
There were I few more, but I do not recall them right now.
You can see by checking ever so often, that many of the subregions are 'revised' every few weeks (or less):
ftp://nhdftp.usgs.gov/SubRegions/High/They are aware of some of these issues. In April their stated emphsis was on a seamless data join with Canada and Mexico; moving the NHD toward 1:4,800 resolution; and modifying the data base structure to handle some of the issues between NHD and GNIS (data base by Sept, but actual data later / much later).
Hope this helps as it is hard to determine how much to say to clarify versus being too detailed and confusing.
BTW - which version of Census were you using and what area are you authoring?
OZ made the mapset pics for my state maps and I believe only one has streams running over topo features - that was due to 'bad' NED data which USGS replaced after 18 months because of user complaints (the pic was not redone)