Investigate why tracy instrumention is unstable in high throughput environments

[marta-lokhova]

Running max TPS tests at ~3600, I ran into some issues capturing data with Tracy

• Zone names show as "???" when attaching to running pods

  • Even with -DTRACY_ON_DEMAND compiled in, zones registered before client connection show as "???"
  • I verified that symbols are present (visible when connecting "at the right time")

• kubectl port-forward randomly disconnects due to unknown reasons. This does not occur at lower TPS rates.