While there's a limit in the amount of (highly toxic) TEA/TEB hypergolic fuel that's carried in the ignition system I'm not sure it's the bottleneck. The limit

While there's a limit in the amount of (highly toxic) TEA/TEB hypergolic fuel that's carried in the ignition system I'm not sure it's the bottleneck. The limited number of times the Merlin 1D can be relit is probably referring to the re-usability aspect (from memory they tested "cycling" the engine something like 40 times in qualification testing, but it's a little unclear exactly what a cycle entails). It's the associated "coking" of the engine with hot and sticky carbon/soot from the incomplete combustion of the kerosene that's been speculated to be a major hinderance to the long-term engines survivability. Other than consumables, it's probably the high (and variable) latency in the switching time that restarting the engine in quick succession that would make it problematic, in addition to the engine damaging aspects you mentioned. Hover-slamming a first-stage with a Merlin 1D has been heavily tested by F9R-Dev1/2 - it's a completely workable solution that minimizes engine restarts over the flight profile to 3 (boostback, re-entry burn, hoverslam). While the switching the engine may be made to be workable solution which can effectively throttle the engines between 0% and 70%, a hoverslam with a vehicle with thrust/weight ratio > 1 gives less stress on the entire (~$40 million F9R first stage) vehicle to be reused.