Ezoic doesn't control ePMV of sites - it's a reporting metric based on the traffic quality and ad bidding process for each individual site. See here: https://support.ezoic.com/kb/article/what-is-epmv
It seems there's a suggestion here that we've played an active role in cutting your earnings potential which simply isn't the case - it wouldn't be possible, and even if it was it wouldn't be beneficial to our company for publishers to earn less.
What Harry and Ethan are getting at as far as user experience / speed / caching changes is concerned is that it could result in your traffic not engaging with the content in the same way, which would drive down the ePMV and earnings.
I think the focus needs to be on the traffic quality differences between months. Comparing a high volume traffic day where users are highly engaged with the content on the site isn't going to be a fair comparison to 'normal' traffic, and we'd expect the ePMV to be higher on those days.
Let's compare two months that don't look to have had a spike in traffic, September compared to August in this instance.
https://analytics.ezoic.com/reports/traffic/trafficDaily
Visits and pageviews increased, but the quality of the traffic decreased --- engaged pageviews / visit, engaged time per visit, and pageviews per visit all decreased. Traffic from low-earning locations also increased in those periods, and while traffic increased from high paying locations as well, their engagement decreased.
The above, combined with the fact that there were no 'prime day bumps' and the UX / speed changes you made, seem to point to why your ePMV hasn't increased lately.