The most important aspect is that the ground thermal is the cause of the cumulus cloud, not some “cloud thermal”, and this is not correctly modelled now in MSFS and creates confusion to RL glider pilots.
The aspect if the direct connection between ground and cloud exists or not is not that important - it is just a matter how long ground source was able to generate rising air, how fast it rises and how low condensation height is - if generation time was short it generated short puff that drifted with the wind (but often on a predictable path!) from the heat source, if the generation time was long it generated long skewed by the wind column of rising airmass that literally connects ground to the forming cloud - this cloud will be elongated horizontally in the downwind direction.
In some places/conditions the first case is more often in other places/conditions the second one, thus in literature both models (“bubble” and “column”) are discussed, like this illustration to a gliding book:
Of course cloud drifts with the wind further and lives until all rising (and drifting) air from the ground thermal reaches condensation height + some sucking due to latent heat, until it slowly dissipates by mixing with surrounding air.
Even on your picture treating this picture literally you have there not a short puff of thermal, but a long thermal column - at picture “a” already almost reching condensation height, while still generating rising air at the ground, so there would be a state in time (lets call it a+) between picture “a” and “b” where the first rising air reach the condensation height, thus Cu cloud is forming, while still at the same time ground source still generates rising air:
There is obviously no mechanism that could prohibit such state. It is easy to calculate that with conditions like: 15min of thermal source lifting air verticaly rising at 3m/s (circling glider would experience it as 2m/s) and condensation height 1500m - as long as 400sec will pass with direct ground-cloud connection - from first lifted airmass to the end of lift generation at ground source (of course could will live longer as this end if airmas still needs another 500sec to rise to confenasation height and longer due to latent heat still sucking some air and slow mixing).
I think with “cloud thermals” Asobo tried to model the lift generated due to latent heat of condensation strong under stormy, strongly vertically evolved clouds, but very gentle and only near the cloud base under typical Cu cloud - and if that is the case it is okayish as it is, the thing requiring fixing are those “ground thermals”.
That is why I think the current model cannot be fixed by only adjustements of parameters - it needs correct modeling of ground thermals generating airmass that in turn should form Cu clouds when reaching condensation height.