In theory, a land value tax (e.g. Georgist single tax), works by the government replacing property taxes with leases on land, as an incentive to invest in property improvements. This obviously works when the owner of the improvement is the holder of the lease on the improved land. However, if the land value increases, and the rent on the lease does not, the holder of the lease, in effect, privatizes the positive network externalities of others' improvements. Therefore, it is necessary that the lease not fix the rent payments for the duration of the lease and, instead, dynamically adjust them to collect (thence distribute as a social good) the increased land value.
At some point a lease holder may be unable or unwilling to pay the rent, causing a lease transfer. If his improvements are mobile, he can pull up stakes and move them. However buildings are typically fixed improvements. Hence there can occur a decoupling of ownership of a fixed improvement from the ownership of the lease on the underlying land. This would appear to be a theoretic problem with the LVT, since the new owner of the lease's right to improve the property would be constrained by the prior owner's use of the land in the form of fixed improvements.
How does LVT theory deal with this decoupling of land lease from improvement ownership with land lease transfer in the absence of improvement ownership transfer?