The added complexity that you bring to the table does not make worth trying to install package by package (or small group of packages)
Depending on the size of the infra-structure, it might be worth:
- Using a local mirror from your own country, where you have much better connectivity, if it exists;
- using a generic proxy - Squid;
- Having a specialized proxy for packages (satellite for RH derivate systems or
apt-cacher-ng for Debian-based); in that way, only one machine does actually pull the package from the Internet and the other pull from the cache );
- For bigger infra-structures, it might be work having a local repository mirror of the distribution.
In some Asian/African countries also connectivity quality to the outside/bad local DNSes can be mitigated using international VPN services. (quite long to explain the whys)
However, being things what they are, there are no miracles solving local bandwidth overprovisioning/exaustion, especially where wireless connections are involved and/or domestic connections.
TLDR You cannot bend the laws of Physics with Unix wizardry alone.
However, if talking about enterprise setups with dedicated/"better" connectivity, going the extra mile to setup a proper intra-structure and fine-tune network and systems configuration does help on getting a slightly better service. However, there are whole books written about the subject, and it is a overly broad subject to be on-topic here.