Hydraulic fractures occur naturally during ice calving events in glaciers, the sudden draining of glacial lakes, the formation of magma-driven dykes and sills, and the failure of dams. Hydraulic fractures are also engineered by injecting a viscous fluid into rock to increase hydro-carbon recovery, for enhanced geothermal energy production, to remediate and dispose of waste water, for preconditioning and cave inducement in mining operations, and, more recently for the generation of lens-shaped fractures for the geomechanical storage of energy. In these engineering applications, once injection is stopped the wellbore is frequently shut-in by closing a valve. Models of hydraulic fractures involve the solution of a set of coupled, degenerate, elasto-hydrodynamic integro-partial differential equations posed on a domain with a singular free boundary. Over the last five decades, considerable research has been dedicated to building accurate models of propagating hydraulic fractures, but very little research has been done on receding hydraulic fractures - in which the fractures are assumed to close on previously created solid surfaces at the tip. In this talk I will discuss recent research [1-4] into the deflation dynamics of hydraulic fractures as they leak fluid to the porous elastic medium after shut-in. This includes a detailed multiscale asymptotic analysis of the tip behaviour of a hydraulic fracture in a permeable elastic medium during arrest and recession. Numerical schemes constructed from these asymptotic solutions has enabled us to establish the emergence of a self-similar solution as the hydraulic fracture approaches collapse. We demonstrate that this so-called Sunset Solution is caused by a fundamental decoupling of kinematics from dynamics in the governing equations, and that this decoupling opens the possibility of characterizing the porous medium by measuring the declining fracture aperture.