All AWS BGP connections dropped due router id changed after unrelated OSPF neighbor established

Hi community stars,

Can anyone think of why after bringing a new OSPF neighbor establishment would cause the router ID change and all my AWS BGP neighborships to all die? 

e.g.

2020:08:19-14:46:51 utm-1 bgpd[17266]: %ADJCHANGE: neighbor 169.254.20.221 Down Router ID changed

After it happened and I bought down the OSPF neighborship everything stablised (it was all flapping before that).  Normally when flaps happen and you lose BGP by bringing up an IGP it would mean that underlying route to your neighbor is blackholed, but the AWS neighbors are all the 169.254 range and we certainly don't have them anywhere else.

Strangely we already have OSPF running, all I was doing was creating a new link downstream across a /29 interconnect and learning a new /24 in the 10s network that doesn't exist anywhere else.

We use the AWS-VPC point and click setup under Site-to-site but I can't see the underlying configuration anywhere, even under printable config the section is blank except for one line that says its turned on. Just this line appears - Amazon VPC status switch (status) = 1

BGP is turned off under Interfaces & Routing so no idea how the router-id is being set.