Put cache on NFS
Currently, the cache is local to each Jenkins node (we have at least 6 of them for GNU/Linux). Building the whole of Qt (including qtwebengine) on 3 cores (otherwise it risks being OOM killed) takes more than one hour, for each distribution supported (qtwebengine cannot be linked statically). That's more than 10 potential builds per machine, times the number of number of machines we have (60), e.g. more than 60 builds of Qt or more than 60 hours of time.
Having the resulting cached builds on a shared NFS mount would improve things greatly for subsequent builds.
As usual with NFS, we'd have to ensure each job writing to it uses
flock to prevent conflicts.