Hey guys,
with my team we're approaching native (only Mgid for now) and testing things out.
I'd like to hear your opinion on this.

A few questions:
1. MGID seems to have an especially higher suspicious visitor/clicks ratio compared to other native traffic sources, and in most cases - Yes, i suspect these are bots.
You could try asking for a refund for sure, however MGID specifically are a bit difficult with refunds. (I've managed to get multiple partial refunds in the past, but considering the time and effort spent arguing and negotiating with them about it, I'm not positive it was even worth it)
2. Some high suspicious widget still seem to perform quite well, so I have personally stopped considering these elements for my blacklisting procedures.
3. Considering your custom conversion names, I assume you're running Crypto? In that case I'd say a $35 is a quite high lead cost even for tier1 GEOs. (MGID's CR will most likely be much lower than that of FB or G, if that's where you're coming from)
I don't consider costs/CPA so much when deciding to block widgets on MGID, since you can use coefficients to decrease the bid and continue running the widget on lower CPC instead of completely blocking it.
What I do consider is the count of visits. For example if visit count is higher than 200 and LP CTR is below 5% OR CVR is below 0.05%, then it likely won't work no matter how low the CPC will be and I might as well block it completely.