We use the ages of old astrophysical objects (OAO) in the redshift range $0
\lesssim z \lesssim 8$ as stringent tests of the late-time cosmic expansion
history. Since the age of the Universe at any redshift is inversely
proportional to $H_0$, requiring that the Universe be older than the oldest
objects it contains at any redshift, provides an upper limit on $H_0$. Using a
combination of galaxies imaged from the CANDELS program and various high-$z$
quasars, we construct an age-redshift diagram of $\gtrsim 100$ OAO up to $z
\sim 8$. Assuming the $\Lambda$CDM model at late times, we find the
95\%~confidence level upper limit $H_0<73.2\,{\rm km}/{\rm s}/{\rm Mpc}$, in
slight disagreement with a host of local $H_0$ measurements. Taken at face
value, and assuming that the OAO ages are reliable, this suggests that
ultimately a combination of pre- and post-recombination ($z \lesssim 10$) new
physics might be required to reconcile cosmic ages with early-time and local
$H_0$ measurements. In the context of the Hubble tension, our results motivate
the study of either combined global pre- and post-recombination modifications
to $\Lambda$CDM, or local new physics which only affects the local $H_0$
measurements.