That's right - during last couple days we have done some major and dramatic changes to our network's perimeter, processing power and "garbage" utilization while allowing our SBCs' CPU power to do the job they were designed to do).
All those measures together allowed us to reduce number of SRV records per returned result thus going back into "unbuggy" (if I can say it this way - like "undead"
) area of all those UAs which stopped performing when caused to use TCP for larger buffers which in turn caused many devices to reboot.
No need to use any specific DN servers anymore - anything goes - any DNS settings, any "A" or "SRV" settings, any devices may go back to older setting or keep whatever worked in past or works now.
I may only still recommend to use SRV settings or at least try to configure them because of above, earlier listed reasons related to specific "features" which may be used (and already used by our implementation) by providers for great increase in reliability, resiliency thus catering to better security and [what else?] - geographical and local redundancy.