Description of problem:
When synchronizing either of two Grafana repos (they are very similar, afaik), memory usage of `pulpcore-worker` is high, up to 6GB. Moreover the usage grew since 6.14 by 10% so a kind of UX regression (previously, Capsule sync of many CVs containing Grafana repos worked still well, now the higher memory demand triggersOOM killer).

The memory consumption might be appropriate as well as the increase since 6.14 - the Grafana repos have big repodata. But since the mem.usage is so huge and there is an evident increase in 6.14->6.15 upgrade, I would like to have some more formal confirmation "it is still ok" (but an improvement in memory consumption is preferred, of course).
 
The two Grafana repos:

https://rpm.grafana.com/
https://packages.grafana.com/oss/rpm

How reproducible:
100%
 

Is this issue a regression from an earlier version:
yes? since memory usage bigger by 10% when syncing the repo
 

Steps to Reproduce:
1. Run on various Satellite versions below script:

hammer product delete --organization-id 1 --name Grafana_03919224_product # this will fail in 1st run, that's OK

hammer product create --organization-id 1 --name Grafana_03919224_product

hammer repository create --organization-id 1 --product Grafana_03919224_product --name Grafana_03919224_repo --content-type yum --download-policy on_demand --url https://rpm.grafana.com/

while true; do date; ps aux | grep -v grep | grep pulpcore-worker; sleep 1; done > pulpcore-worker.ps.grafana.log &
pid=$!
sleep 5

repoid=$(hammer repository list | grep Grafana_03919224_repo | awk '{ print $1 }')
hammer repository sync --organization-id 1 --product Grafana_03919224_product --id $repoid
sleep 5
kill $pid

sort -nk6 pulpcore-worker.ps.grafana.log | tail

2. Check the output of last line - biggest memory consumption of a pulp worker over the time.

3. Re-run the script once again. It will delete the product and recreate everything again. And check memory usage again - dunno why but second (and any subsequent) repo sync does consume much more memory - on either Satellite version.

Actual behavior:
Comparison of memory usage on different Satellite versions:

https://rpm.grafana.com:
FIRST SYNC:
6.12: 4970472 = 4.74 GiB
6.13: 5083524 = 4.85 GiB
6.14: 4225924 = 4.03 GiB
6.15: 4675268 = 4.46 GiB

NEXT SYNC:
6.12: 5434556 = 5.18 GiB
6.13: 5510888 = 5.26 GiB
6.14: 5518836 = 5.26 GiB
6.15: 6006344 = 5.73 GiB
6.16: 5836332 = 5.57 GiB

https://packages.grafana.com/oss/rpm:
FIRST SYNC: (havent measured)

NEXT SYNC:
6.12: 5440196 = 5.19 GiB
6.13: 5428680 = 5.17 GiB
6.14: 5435956 = 5.18 GiB
6.15: 6017092 = 5.74 GiB
6.16: 5862764 = 5.60 GiB

Expected behavior:
"Lower memory usage". Hard to say what is the reasonable value, but at least the mem.usage should be on the 6.12-6.14 level (see the increase in 6.15).

Business Impact / Additional info:
During Capsule sync, 4 or 8 pulpcore workers require 6GB of RAM each when syncing the Grafana repos present in multiple versions of several CVs. That requires more memory than the Capsule is recommended to have (as a minimum).