I'm only aware of the ESA and SpaceX incident in 2019. Does this happen more regularly especially now with more active Starlink satellites and other constellations?
More importantly, are there efforts to track these types of near collisions where one or both agencies react (or choose not to) to avoid collisions?
Well, if we want to talk about SpaceX and Starlink, they're obligated by regulation to provide semi annual reports on the operations of Starlink. You can find the reports here (as well as a hilarious amount of very snide back and forth between industry on regulatory approval of Starlink). In the period of 2021 June 1 (~1700 sats) to 2021 November 30 (~1800 sats) they conducted 3,333 maneuvers, although this does number include maneuvers avoiding debris, not just satellite v satellite. There was also 10 occurrences where other operators asked SpaceX not to perform a maneuver because those operators preferred to maneuver their satellites.
Per the most recent (March 2022) Orbital Debris Quarterly News issued by NASA, the International Space Station has conducted 30 orbital debris avoidance maneuvers since 1999. That's a bit over one per year, on average.
The ISS is certainly going to be an outlier with regard to the frequency at which it performs collision avoidance maneuvers, for several reasons:
The first three of the above reasons means NASA and Roscosmos have to be very aggressive regarding avoiding potential collisions with debris. The final reason means that NASA and Roscosmos can afford to be very aggressive with regard to avoiding potential collisions. While other spacecraft are expensive, none come close to the value of the ISS. Most spacecraft are not refuelable, so most space operators have to tolerate more risk than the ISS. I suspect that for most spacecraft in low Earth orbit that it's well under one collision avoidance maneuver per year.
The Space Data Association (SDA) has, as a part of its reason for being, the role of sharing good quality orbit ephemeris amongst subscribing satellite operators.
Operators own ranging of their satellites is usually rather more accurate than the publically available TLEs. The SDA also have some link (I don't know the details) with JSPOC to access better quality ephemeris than TLEs for uncontrolled objects or those of non-subscribing operators.
How often is a manoeuvre actually a) necessary in one person's view or b) actually performed? I don't think anyone has the big picture - there are ball park estimates in the debris community, its a way of spreading a message about the cost of space debris.
